totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
# define OLD_NTDOMAIN 1
/*
Unix SMB / Netbios implementation .
Version 3.0
change notify handling - linux kernel based implementation
Copyright ( C ) Andrew Tridgell 2000
This program is free software ; you can redistribute it and / or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation ; either version 2 of the License , or
( at your option ) any later version .
This program is distributed in the hope that it will be useful ,
but WITHOUT ANY WARRANTY ; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE . See the
GNU General Public License for more details .
You should have received a copy of the GNU General Public License
along with this program ; if not , write to the Free Software
Foundation , Inc . , 675 Mass Ave , Cambridge , MA 0213 9 , USA .
*/
# include "includes.h"
# if HAVE_KERNEL_CHANGE_NOTIFY
extern int DEBUGLEVEL ;
2000-06-19 21:30:27 +00:00
static VOLATILE SIG_ATOMIC_T fd_pending ;
static VOLATILE SIG_ATOMIC_T signals_received ;
static VOLATILE SIG_ATOMIC_T signals_processed ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
# ifndef DN_ACCESS
# define DN_ACCESS 0x00000001 /* File accessed in directory */
# define DN_MODIFY 0x00000002 /* File modified in directory */
# define DN_CREATE 0x00000004 /* File created in directory */
# define DN_DELETE 0x00000008 /* File removed from directory */
# define DN_RENAME 0x00000010 /* File renamed in directory */
2000-06-13 09:31:43 +00:00
# define DN_ATTRIB 0x00000020 /* File changed attribute */
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
# define DN_MULTISHOT 0x80000000 /* Don't remove notifier */
# endif
# ifndef RT_SIGNAL_NOTIFY
# define RT_SIGNAL_NOTIFY 34
# endif
2000-06-12 17:06:00 +00:00
# ifndef F_SETSIG
# define F_SETSIG 10
# endif
# ifndef F_NOTIFY
# define F_NOTIFY 1026
# endif
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
/****************************************************************************
This is the structure to keep the information needed to
determine if a directory has changed .
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
struct change_data {
int directory_handle ;
} ;
/****************************************************************************
the signal handler for change notify
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
static void signal_handler ( int signal , siginfo_t * info , void * unused )
{
BlockSignals ( True , signal ) ;
2000-06-19 21:30:27 +00:00
fd_pending = ( SIG_ATOMIC_T ) info - > si_fd ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
signals_received + + ;
sys_select_signal ( ) ;
}
/****************************************************************************
check if a change notify should be issued
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
static BOOL kernel_check_notify ( connection_struct * conn , uint16 vuid , char * path , uint32 flags , void * datap , time_t t )
{
struct change_data * data = ( struct change_data * ) datap ;
2000-06-19 21:30:27 +00:00
if ( data - > directory_handle ! = ( int ) fd_pending ) return False ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
2000-06-19 21:30:27 +00:00
DEBUG ( 3 , ( " kernel change notify on %s fd=%d \n " , path , ( int ) fd_pending ) ) ;
2000-06-12 17:06:00 +00:00
2000-06-19 21:30:27 +00:00
close ( ( int ) fd_pending ) ;
fd_pending = ( SIG_ATOMIC_T ) - 1 ;
data - > directory_handle = - 1 ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
signals_processed + + ;
BlockSignals ( False , RT_SIGNAL_NOTIFY ) ;
return True ;
}
/****************************************************************************
remove a change notify data structure
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
static void kernel_remove_notify ( void * datap )
{
struct change_data * data = ( struct change_data * ) datap ;
2000-06-12 17:06:00 +00:00
int fd = data - > directory_handle ;
if ( fd ! = - 1 ) {
2000-06-19 21:30:27 +00:00
if ( fd = = ( int ) fd_pending ) {
fd_pending = ( SIG_ATOMIC_T ) - 1 ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
signals_processed + + ;
BlockSignals ( False , RT_SIGNAL_NOTIFY ) ;
}
2000-06-12 17:06:00 +00:00
close ( fd ) ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
}
free ( data ) ;
2000-06-12 17:06:00 +00:00
DEBUG ( 3 , ( " removed kernel change notify fd=%d \n " , fd ) ) ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
}
/****************************************************************************
register a change notify request
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
static void * kernel_register_notify ( connection_struct * conn , char * path , uint32 flags )
{
struct change_data data ;
int fd ;
unsigned long kernel_flags ;
2000-06-12 17:06:00 +00:00
fd = dos_open ( path , O_RDONLY , 0 ) ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
if ( fd = = - 1 ) {
2000-06-12 17:06:00 +00:00
DEBUG ( 3 , ( " Failed to open directory %s for change notify \n " , path ) ) ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
return NULL ;
}
if ( fcntl ( fd , F_SETSIG , RT_SIGNAL_NOTIFY ) = = - 1 ) {
DEBUG ( 3 , ( " Failed to set signal handler for change notify \n " ) ) ;
return NULL ;
}
2000-06-13 09:31:43 +00:00
kernel_flags = DN_CREATE | DN_DELETE | DN_RENAME ; /* creation/deletion changes everything! */
if ( flags & FILE_NOTIFY_CHANGE_FILE ) kernel_flags | = DN_MODIFY ;
2000-06-12 17:06:00 +00:00
if ( flags & FILE_NOTIFY_CHANGE_DIR_NAME ) kernel_flags | = DN_RENAME | DN_DELETE ;
2000-06-13 09:31:43 +00:00
if ( flags & FILE_NOTIFY_CHANGE_ATTRIBUTES ) kernel_flags | = DN_ATTRIB ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
if ( flags & FILE_NOTIFY_CHANGE_SIZE ) kernel_flags | = DN_MODIFY ;
if ( flags & FILE_NOTIFY_CHANGE_LAST_WRITE ) kernel_flags | = DN_MODIFY ;
if ( flags & FILE_NOTIFY_CHANGE_LAST_ACCESS ) kernel_flags | = DN_ACCESS ;
if ( flags & FILE_NOTIFY_CHANGE_CREATION ) kernel_flags | = DN_CREATE ;
2000-06-13 09:31:43 +00:00
if ( flags & FILE_NOTIFY_CHANGE_SECURITY ) kernel_flags | = DN_ATTRIB ;
if ( flags & FILE_NOTIFY_CHANGE_EA ) kernel_flags | = DN_ATTRIB ;
if ( flags & FILE_NOTIFY_CHANGE_FILE_NAME ) kernel_flags | = DN_RENAME | DN_DELETE ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
if ( fcntl ( fd , F_NOTIFY , kernel_flags ) = = - 1 ) {
DEBUG ( 3 , ( " Failed to set async flag for change notify \n " ) ) ;
return NULL ;
}
data . directory_handle = fd ;
2000-06-13 06:33:15 +00:00
DEBUG ( 3 , ( " kernel change notify on %s (ntflags=0x%x flags=0x%x) fd=%d \n " ,
path , ( int ) flags , ( int ) kernel_flags , fd ) ) ;
2000-06-12 17:06:00 +00:00
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
return ( void * ) memdup ( & data , sizeof ( data ) ) ;
}
2000-06-12 17:06:00 +00:00
/****************************************************************************
see if the kernel supports change notify
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
static BOOL kernel_notify_available ( void )
{
int fd , ret ;
fd = open ( " /tmp " , O_RDONLY ) ;
if ( fd = = - 1 ) return False ; /* uggh! */
ret = fcntl ( fd , F_NOTIFY , 0 ) ;
close ( fd ) ;
2000-06-14 07:11:27 +00:00
return ret = = 0 ;
2000-06-12 17:06:00 +00:00
}
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
/****************************************************************************
setup kernel based change notify
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
struct cnotify_fns * kernel_notify_init ( void )
{
static struct cnotify_fns cnotify ;
struct sigaction act ;
act . sa_handler = NULL ;
act . sa_sigaction = signal_handler ;
act . sa_flags = SA_SIGINFO ;
if ( sigaction ( RT_SIGNAL_NOTIFY , & act , NULL ) ! = 0 ) {
DEBUG ( 0 , ( " Failed to setup RT_SIGNAL_NOTIFY handler \n " ) ) ;
return NULL ;
}
2000-06-14 07:11:27 +00:00
if ( ! kernel_notify_available ( ) ) return NULL ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
cnotify . register_notify = kernel_register_notify ;
cnotify . check_notify = kernel_check_notify ;
cnotify . remove_notify = kernel_remove_notify ;
2000-06-14 07:11:27 +00:00
cnotify . select_time = - 1 ;
totally rewrote the async signal, notification and oplock notification
handling in Samba. This was needed due to several limitations and
races in the previous code - as a side effect the new code is much
cleaner :)
in summary:
- changed sys_select() to avoid a signal/select race condition. It is a
rare race but once we have signals doing notification and oplocks it
is important.
- changed our main processing loop to take advantage of the new
sys_select semantics
- split the notify code into implementaion dependent and general
parts. Added the following structure that defines an implementation:
struct cnotify_fns {
void * (*register_notify)(connection_struct *conn, char *path, uint32 flags);
BOOL (*check_notify)(connection_struct *conn, uint16 vuid, char *path, uint32 flags, void *data, time_t t);
void (*remove_notify)(void *data);
};
then I wrote two implementations, one using hash/poll (like our old
code) and the other using the new Linux kernel change notify. It
should be easy to add other change notify implementations by creating
a sructure of the above type.
- fixed a bug in change notify where we were returning the wrong error
code.
- rewrote the core change notify code to be much simpler
- moved to real-time signals for leases and change notify
Amazingly, it all seems to work. I was very surprised!
(This used to be commit 44766c39e0027c762bee8b33b12c621c109a3267)
2000-06-12 15:53:31 +00:00
return & cnotify ;
}
# else
void notify_kernel_dummy ( void ) { }
# endif /* HAVE_KERNEL_CHANGE_NOTIFY */
# undef OLD_NTDOMAIN
