minix/servers/vfs/select.c

709 lines
20 KiB
C
Raw Normal View History

/* Implement entry point to select system call.
*
* The entry points into this file are
* do_select: perform the SELECT system call
* select_callback: notify select system of possible fd operation
* select_notified: low-level entry for device notifying select
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
* select_unsuspend_by_endpt: cancel a blocking select on exiting driver
*
* Changes:
* 6 june 2005 Created (Ben Gras)
*/
2005-07-27 16:30:25 +02:00
#define DEBUG_SELECT 0
#include "fs.h"
#include "select.h"
#include "file.h"
#include "vnode.h"
#include <sys/time.h>
#include <sys/select.h>
#include <minix/com.h>
#include <minix/u64.h>
#include <string.h>
/* max. number of simultaneously pending select() calls */
#define MAXSELECTS 25
PRIVATE struct selectentry {
struct fproc *requestor; /* slot is free iff this is NULL */
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
int req_endpt;
fd_set readfds, writefds, errorfds;
fd_set ready_readfds, ready_writefds, ready_errorfds;
fd_set *vir_readfds, *vir_writefds, *vir_errorfds;
struct filp *filps[OPEN_MAX];
int type[OPEN_MAX];
int nfds, nreadyfds;
clock_t expiry;
timer_t timer; /* if expiry > 0 */
} selecttab[MAXSELECTS];
FORWARD _PROTOTYPE(int select_reevaluate, (struct filp *fp));
2005-08-29 18:47:18 +02:00
FORWARD _PROTOTYPE(int select_request_file,
(struct filp *f, int *ops, int block));
FORWARD _PROTOTYPE(int select_match_file, (struct filp *f));
2005-08-29 18:47:18 +02:00
FORWARD _PROTOTYPE(int select_request_general,
(struct filp *f, int *ops, int block));
FORWARD _PROTOTYPE(int select_major_match,
(int match_major, struct filp *file));
FORWARD _PROTOTYPE(void select_cancel_all, (struct selectentry *e));
FORWARD _PROTOTYPE(void select_wakeup, (struct selectentry *e, int r));
FORWARD _PROTOTYPE(void select_return, (struct selectentry *, int));
/* The Open Group:
* "The pselect() and select() functions shall support
* regular files, terminal and pseudo-terminal devices,
* STREAMS-based files, FIFOs, pipes, and sockets."
*/
PRIVATE struct fdtype {
int (*select_request)(struct filp *, int *ops, int block);
int (*select_match)(struct filp *);
int select_major;
} fdtypes[] = {
{ select_request_file, select_match_file, 0 },
{ select_request_general, NULL, TTY_MAJOR },
{ select_request_general, NULL, INET_MAJOR },
{ select_request_pipe, select_match_pipe, 0 },
{ select_request_general, NULL, LOG_MAJOR },
};
#define SEL_FDS (sizeof(fdtypes) / sizeof(fdtypes[0]))
/* Open Group:
* "File descriptors associated with regular files shall always select true
* for ready to read, ready to write, and error conditions."
*/
/*===========================================================================*
* select_request_file *
*===========================================================================*/
PRIVATE int select_request_file(struct filp *f, int *ops, int block)
{
/* output *ops is input *ops */
return SEL_OK;
}
/*===========================================================================*
* select_match_file *
*===========================================================================*/
PRIVATE int select_match_file(struct filp *file)
{
if (file && file->filp_vno && (file->filp_vno->v_mode & I_REGULAR))
return 1;
return 0;
}
/*===========================================================================*
* select_request_general *
*===========================================================================*/
PRIVATE int select_request_general(struct filp *f, int *ops, int block)
{
2005-08-25 14:30:43 +02:00
int rops = *ops;
2005-09-11 18:45:46 +02:00
if (block) rops |= SEL_NOTIFY;
*ops = dev_io(DEV_SELECT, f->filp_vno->v_sdev, rops, NULL, cvu64(0),
0, 0);
2005-09-11 18:45:46 +02:00
if (*ops < 0)
return SEL_ERR;
return SEL_OK;
}
/*===========================================================================*
* select_major_match *
*===========================================================================*/
PRIVATE int select_major_match(int match_major, struct filp *file)
{
int major;
if (!(file && file->filp_vno &&
(file->filp_vno->v_mode & I_TYPE) == I_CHAR_SPECIAL))
return 0;
major = (file->filp_vno->v_sdev >> MAJOR) & BYTE;
2005-09-11 18:45:46 +02:00
if (major == match_major)
2005-06-28 16:59:00 +02:00
return 1;
return 0;
}
2005-09-22 23:17:22 +02:00
/*===========================================================================*
* tab2ops *
*===========================================================================*/
PRIVATE int tab2ops(int fd, struct selectentry *e)
{
return (FD_ISSET(fd, &e->readfds) ? SEL_RD : 0) |
(FD_ISSET(fd, &e->writefds) ? SEL_WR : 0) |
(FD_ISSET(fd, &e->errorfds) ? SEL_ERR : 0);
}
2005-09-22 23:17:22 +02:00
/*===========================================================================*
* ops2tab *
*===========================================================================*/
PRIVATE void ops2tab(int ops, int fd, struct selectentry *e)
{
2005-09-11 18:45:46 +02:00
if ((ops & SEL_RD) && e->vir_readfds && FD_ISSET(fd, &e->readfds)
&& !FD_ISSET(fd, &e->ready_readfds)) {
FD_SET(fd, &e->ready_readfds);
e->nreadyfds++;
}
2005-09-11 18:45:46 +02:00
if ((ops & SEL_WR) && e->vir_writefds && FD_ISSET(fd, &e->writefds)
&& !FD_ISSET(fd, &e->ready_writefds)) {
FD_SET(fd, &e->ready_writefds);
e->nreadyfds++;
}
2005-09-11 18:45:46 +02:00
if ((ops & SEL_ERR) && e->vir_errorfds && FD_ISSET(fd, &e->errorfds)
&& !FD_ISSET(fd, &e->ready_errorfds)) {
FD_SET(fd, &e->ready_errorfds);
e->nreadyfds++;
}
return;
}
2005-09-22 23:17:22 +02:00
/*===========================================================================*
* copy_fdsets *
*===========================================================================*/
PRIVATE void copy_fdsets(struct selectentry *e)
{
int fd_setsize;
if(e->nfds < 0 || e->nfds > OPEN_MAX)
panic(__FILE__, "select copy_fdsets: e->nfds wrong", e->nfds);
/* Only copy back as many bits as the user expects. */
fd_setsize = _FDSETWORDS(e->nfds)*_FDSETBITSPERWORD/8;
if (e->vir_readfds)
sys_vircopy(SELF, D, (vir_bytes) &e->ready_readfds,
e->req_endpt, D, (vir_bytes) e->vir_readfds, fd_setsize);
if (e->vir_writefds)
sys_vircopy(SELF, D, (vir_bytes) &e->ready_writefds,
e->req_endpt, D, (vir_bytes) e->vir_writefds, fd_setsize);
if (e->vir_errorfds)
sys_vircopy(SELF, D, (vir_bytes) &e->ready_errorfds,
e->req_endpt, D, (vir_bytes) e->vir_errorfds, fd_setsize);
return;
}
/*===========================================================================*
2005-09-11 18:45:46 +02:00
* do_select *
*===========================================================================*/
PUBLIC int do_select(void)
{
int r, nfds, is_timeout = 1, nonzero_timeout = 0,
fd, s, block = 0, fd_setsize;
struct timeval timeout;
nfds = m_in.SEL_NFDS;
if (nfds < 0 || nfds > OPEN_MAX)
return EINVAL;
for(s = 0; s < MAXSELECTS; s++)
2005-09-11 18:45:46 +02:00
if (!selecttab[s].requestor)
break;
2005-09-11 18:45:46 +02:00
if (s >= MAXSELECTS)
return ENOSPC;
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
selecttab[s].req_endpt = who_e;
selecttab[s].nfds = 0;
selecttab[s].nreadyfds = 0;
memset(selecttab[s].filps, 0, sizeof(selecttab[s].filps));
/* defaults */
FD_ZERO(&selecttab[s].readfds);
FD_ZERO(&selecttab[s].writefds);
FD_ZERO(&selecttab[s].errorfds);
FD_ZERO(&selecttab[s].ready_readfds);
FD_ZERO(&selecttab[s].ready_writefds);
FD_ZERO(&selecttab[s].ready_errorfds);
selecttab[s].vir_readfds = (fd_set *) m_in.SEL_READFDS;
selecttab[s].vir_writefds = (fd_set *) m_in.SEL_WRITEFDS;
selecttab[s].vir_errorfds = (fd_set *) m_in.SEL_ERRORFDS;
/* Copy args. Our storage size is zeroed above. Only copy
* as many bits as user has supplied (nfds).
* Could be compiled with a different OPEN_MAX or FD_SETSIZE.
* If nfds is too large, we have already returned above.
*/
fd_setsize = _FDSETWORDS(nfds)*_FDSETBITSPERWORD/8;
2005-09-11 18:45:46 +02:00
if (selecttab[s].vir_readfds
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
&& (r=sys_vircopy(who_e, D, (vir_bytes) m_in.SEL_READFDS,
SELF, D, (vir_bytes) &selecttab[s].readfds, fd_setsize)) != OK)
return r;
2005-09-11 18:45:46 +02:00
if (selecttab[s].vir_writefds
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
&& (r=sys_vircopy(who_e, D, (vir_bytes) m_in.SEL_WRITEFDS,
SELF, D, (vir_bytes) &selecttab[s].writefds, fd_setsize)) != OK)
return r;
2005-09-11 18:45:46 +02:00
if (selecttab[s].vir_errorfds
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
&& (r=sys_vircopy(who_e, D, (vir_bytes) m_in.SEL_ERRORFDS,
SELF, D, (vir_bytes) &selecttab[s].errorfds, fd_setsize)) != OK)
return r;
2005-09-11 18:45:46 +02:00
if (!m_in.SEL_TIMEOUT)
is_timeout = nonzero_timeout = 0;
else
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
if ((r=sys_vircopy(who_e, D, (vir_bytes) m_in.SEL_TIMEOUT,
SELF, D, (vir_bytes) &timeout, sizeof(timeout))) != OK)
return r;
/* No nonsense in the timeval please. */
2005-09-11 18:45:46 +02:00
if (is_timeout && (timeout.tv_sec < 0 || timeout.tv_usec < 0))
return EINVAL;
/* if is_timeout if 0, we block forever. otherwise, if nonzero_timeout
* is 0, we do a poll (don't block). otherwise, we block up to the
* specified time interval.
*/
2005-09-11 18:45:46 +02:00
if (is_timeout && (timeout.tv_sec > 0 || timeout.tv_usec > 0))
nonzero_timeout = 1;
2005-09-11 18:45:46 +02:00
if (nonzero_timeout || !is_timeout)
block = 1;
else
block = 0; /* timeout set as (0,0) - this effects a poll */
/* no timeout set (yet) */
selecttab[s].expiry = 0;
for(fd = 0; fd < nfds; fd++) {
int orig_ops, ops, t, type = -1, r;
struct filp *filp;
2005-09-11 18:45:46 +02:00
if (!(orig_ops = ops = tab2ops(fd, &selecttab[s])))
continue;
2005-09-11 18:45:46 +02:00
if (!(filp = selecttab[s].filps[fd] = get_filp(fd))) {
select_cancel_all(&selecttab[s]);
return EBADF;
}
for(t = 0; t < SEL_FDS; t++) {
2005-09-11 18:45:46 +02:00
if (fdtypes[t].select_match) {
if (fdtypes[t].select_match(filp)) {
#if DEBUG_SELECT
printf("select: fd %d is type %d ", fd, t);
#endif
2005-09-11 18:45:46 +02:00
if (type != -1)
printf("select: double match\n");
type = t;
}
2005-09-11 18:45:46 +02:00
} else if (select_major_match(fdtypes[t].select_major, filp)) {
type = t;
}
}
/* Open Group:
* "The pselect() and select() functions shall support
* regular files, terminal and pseudo-terminal devices,
* STREAMS-based files, FIFOs, pipes, and sockets. The
* behavior of pselect() and select() on file descriptors
* that refer to other types of file is unspecified."
*
* If all types are implemented, then this is another
* type of file and we get to do whatever we want.
*/
2005-09-11 18:45:46 +02:00
if (type == -1)
2005-06-28 16:59:00 +02:00
{
#if DEBUG_SELECT
2005-06-28 16:59:00 +02:00
printf("do_select: bad type\n");
#endif
return EBADF;
2005-06-28 16:59:00 +02:00
}
selecttab[s].type[fd] = type;
2005-09-11 18:45:46 +02:00
if ((selecttab[s].filps[fd]->filp_select_ops & ops) != ops) {
int wantops;
/* Request the select on this fd. */
#if DEBUG_SELECT
printf("%p requesting ops %d -> ",
selecttab[s].filps[fd],
selecttab[s].filps[fd]->filp_select_ops);
#endif
wantops = (selecttab[s].filps[fd]->filp_select_ops |= ops);
#if DEBUG_SELECT
printf("%d\n", selecttab[s].filps[fd]->filp_select_ops);
#endif
2005-09-11 18:45:46 +02:00
if ((r = fdtypes[type].select_request(filp,
&wantops, block)) != SEL_OK) {
/* error or bogus return code.. backpaddle */
select_cancel_all(&selecttab[s]);
printf("select: select_request returned error\n");
return EINVAL;
}
2005-09-11 18:45:46 +02:00
if (wantops) {
if (wantops & ops) {
/* operations that were just requested
* are ready to go right away
*/
ops2tab(wantops, fd, &selecttab[s]);
}
/* if there are any other select()s blocking
* on these operations of this fp, they can
* be awoken too
*/
select_callback(filp, ops);
}
#if DEBUG_SELECT
printf("select request ok; ops returned %d\n", wantops);
#endif
} else {
#if DEBUG_SELECT
printf("select already happening on that filp\n");
#endif
}
selecttab[s].nfds = fd+1;
selecttab[s].filps[fd]->filp_selectors++;
#if DEBUG_SELECT
printf("[fd %d ops: %d] ", fd, ops);
#endif
}
2005-09-11 18:45:46 +02:00
if (selecttab[s].nreadyfds > 0 || !block) {
/* fd's were found that were ready to go right away, and/or
* we were instructed not to block at all. Must return
* immediately.
*/
copy_fdsets(&selecttab[s]);
select_cancel_all(&selecttab[s]);
selecttab[s].requestor = NULL;
/* Open Group:
* "Upon successful completion, the pselect() and select()
* functions shall return the total number of bits
* set in the bit masks."
*/
#if DEBUG_SELECT
printf("returning\n");
#endif
return selecttab[s].nreadyfds;
}
#if DEBUG_SELECT
printf("not returning (%d, %d)\n", selecttab[s].nreadyfds, block);
#endif
/* Convert timeval to ticks and set the timer. If it fails, undo
* all, return error.
*/
2005-09-11 18:45:46 +02:00
if (is_timeout) {
int ticks;
/* Open Group:
* "If the requested timeout interval requires a finer
* granularity than the implementation supports, the
* actual timeout interval shall be rounded up to the next
* supported value."
*/
#define USECPERSEC 1000000
while(timeout.tv_usec >= USECPERSEC) {
/* this is to avoid overflow with *HZ below */
timeout.tv_usec -= USECPERSEC;
timeout.tv_sec++;
}
ticks = timeout.tv_sec * HZ +
(timeout.tv_usec * HZ + USECPERSEC-1) / USECPERSEC;
selecttab[s].expiry = ticks;
fs_set_timer(&selecttab[s].timer, ticks, select_timeout_check, s);
#if DEBUG_SELECT
printf("%d: blocking %d ticks\n", s, ticks);
#endif
}
/* if we're blocking, the table entry is now valid. */
selecttab[s].requestor = fp;
/* process now blocked */
suspend(XSELECT);
return SUSPEND;
}
2005-09-22 23:17:22 +02:00
/*===========================================================================*
* select_cancel_all *
*===========================================================================*/
PRIVATE void select_cancel_all(struct selectentry *e)
{
int fd;
for(fd = 0; fd < e->nfds; fd++) {
struct filp *fp;
fp = e->filps[fd];
2005-09-11 18:45:46 +02:00
if (!fp) {
#if DEBUG_SELECT
printf("[ fd %d/%d NULL ] ", fd, e->nfds);
#endif
continue;
}
2005-09-11 18:45:46 +02:00
if (fp->filp_selectors < 1) {
#if DEBUG_SELECT
printf("select: %d selectors?!\n", fp->filp_selectors);
#endif
continue;
}
fp->filp_selectors--;
e->filps[fd] = NULL;
select_reevaluate(fp);
}
2005-09-11 18:45:46 +02:00
if (e->expiry > 0) {
#if DEBUG_SELECT
printf("cancelling timer %d\n", e - selecttab);
#endif
fs_cancel_timer(&e->timer);
e->expiry = 0;
}
return;
}
2005-09-22 23:17:22 +02:00
/*===========================================================================*
* select_wakeup *
*===========================================================================*/
PRIVATE void select_wakeup(struct selectentry *e, int r)
{
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
revive(e->req_endpt, r);
}
2005-09-22 23:17:22 +02:00
/*===========================================================================*
* select_reevaluate *
*===========================================================================*/
PRIVATE int select_reevaluate(struct filp *fp)
{
int s, remain_ops = 0, fd, type = -1;
2005-09-11 18:45:46 +02:00
if (!fp) {
printf("fs: select: reevalute NULL fp\n");
return 0;
}
for(s = 0; s < MAXSELECTS; s++) {
2005-09-11 18:45:46 +02:00
if (!selecttab[s].requestor)
continue;
for(fd = 0; fd < selecttab[s].nfds; fd++)
2005-09-11 18:45:46 +02:00
if (fp == selecttab[s].filps[fd]) {
remain_ops |= tab2ops(fd, &selecttab[s]);
type = selecttab[s].type[fd];
}
}
/* If there are any select()s open that want any operations on
* this fd that haven't been satisfied by this callback, then we're
* still in the market for it.
*/
fp->filp_select_ops = remain_ops;
#if DEBUG_SELECT
printf("remaining operations on fp are %d\n", fp->filp_select_ops);
#endif
return remain_ops;
}
/*===========================================================================*
* select_return *
*===========================================================================*/
PRIVATE void select_return(struct selectentry *s, int r)
{
select_cancel_all(s);
copy_fdsets(s);
select_wakeup(s, r ? r : s->nreadyfds);
s->requestor = NULL;
}
/*===========================================================================*
2005-09-11 18:45:46 +02:00
* select_callback *
*===========================================================================*/
PUBLIC int select_callback(struct filp *fp, int ops)
{
2005-08-25 14:30:43 +02:00
int s, fd, want_ops, type;
/* We are being notified that file pointer fp is available for
* operations 'ops'. We must re-register the select for
* operations that we are still interested in, if any.
*/
want_ops = 0;
type = -1;
for(s = 0; s < MAXSELECTS; s++) {
int wakehim = 0;
2005-09-11 18:45:46 +02:00
if (!selecttab[s].requestor)
continue;
for(fd = 0; fd < selecttab[s].nfds; fd++) {
2005-09-11 18:45:46 +02:00
if (!selecttab[s].filps[fd])
continue;
2005-09-11 18:45:46 +02:00
if (selecttab[s].filps[fd] == fp) {
int this_want_ops;
this_want_ops = tab2ops(fd, &selecttab[s]);
want_ops |= this_want_ops;
2005-09-11 18:45:46 +02:00
if (this_want_ops & ops) {
/* this select() has been satisfied. */
ops2tab(ops, fd, &selecttab[s]);
wakehim = 1;
}
type = selecttab[s].type[fd];
}
}
if (wakehim)
select_return(&selecttab[s], 0);
}
return 0;
}
/*===========================================================================*
2005-09-11 18:45:46 +02:00
* select_notified *
*===========================================================================*/
PUBLIC int select_notified(int major, int minor, int selected_ops)
{
int s, f, t;
#if DEBUG_SELECT
printf("select callback: %d, %d: %d\n", major, minor, selected_ops);
#endif
for(t = 0; t < SEL_FDS; t++)
2005-09-11 18:45:46 +02:00
if (!fdtypes[t].select_match && fdtypes[t].select_major == major)
break;
2005-09-11 18:45:46 +02:00
if (t >= SEL_FDS) {
#if DEBUG_SELECT
printf("select callback: no fdtype found for device %d\n", major);
#endif
return OK;
}
/* We have a select callback from major device no.
* d, which corresponds to our select type t.
*/
for(s = 0; s < MAXSELECTS; s++) {
int s_minor, ops;
2005-09-11 18:45:46 +02:00
if (!selecttab[s].requestor)
continue;
for(f = 0; f < selecttab[s].nfds; f++) {
2005-09-11 18:45:46 +02:00
if (!selecttab[s].filps[f] ||
!select_major_match(major, selecttab[s].filps[f]))
continue;
ops = tab2ops(f, &selecttab[s]);
s_minor =
(selecttab[s].filps[f]->filp_vno->v_sdev >> MINOR)
& BYTE;
2005-09-11 18:45:46 +02:00
if ((s_minor == minor) &&
(selected_ops & ops)) {
select_callback(selecttab[s].filps[f], (selected_ops & ops));
2005-06-28 16:59:00 +02:00
}
}
}
return OK;
}
/*===========================================================================*
* init_select *
*===========================================================================*/
PUBLIC void init_select(void)
{
int s;
for(s = 0; s < MAXSELECTS; s++)
fs_init_timer(&selecttab[s].timer);
}
/*===========================================================================*
2005-09-11 18:45:46 +02:00
* select_forget *
*===========================================================================*/
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
PUBLIC void select_forget(int proc_e)
{
/* something has happened (e.g. signal delivered that interrupts
* select()). totally forget about the select().
*/
int s;
for(s = 0; s < MAXSELECTS; s++) {
2005-09-11 18:45:46 +02:00
if (selecttab[s].requestor &&
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
selecttab[s].req_endpt == proc_e) {
break;
}
}
2005-09-11 18:45:46 +02:00
if (s >= MAXSELECTS) {
#if DEBUG_SELECT
printf("select: cancelled select() not found");
#endif
return;
}
select_cancel_all(&selecttab[s]);
selecttab[s].requestor = NULL;
return;
}
/*===========================================================================*
2005-09-11 18:45:46 +02:00
* select_timeout_check *
*===========================================================================*/
PUBLIC void select_timeout_check(timer_t *timer)
{
2005-08-25 14:30:43 +02:00
int s;
s = tmr_arg(timer)->ta_int;
2005-09-11 18:45:46 +02:00
if (s < 0 || s >= MAXSELECTS) {
#if DEBUG_SELECT
printf("select: bogus slot arg to watchdog %d\n", s);
#endif
return;
}
2005-09-11 18:45:46 +02:00
if (!selecttab[s].requestor) {
#if DEBUG_SELECT
printf("select: no requestor in watchdog\n");
#endif
return;
}
2005-09-11 18:45:46 +02:00
if (selecttab[s].expiry <= 0) {
#if DEBUG_SELECT
printf("select: strange expiry value in watchdog\n", s);
#endif
return;
}
selecttab[s].expiry = 0;
select_return(&selecttab[s], 0);
return;
}
/*===========================================================================*
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
* select_unsuspend_by_endpt *
*===========================================================================*/
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
PUBLIC void select_unsuspend_by_endpt(int proc_e)
{
int fd, s;
for(s = 0; s < MAXSELECTS; s++) {
if (!selecttab[s].requestor)
continue;
for(fd = 0; fd < selecttab[s].nfds; fd++) {
int maj;
if (!selecttab[s].filps[fd] || !selecttab[s].filps[fd]->filp_vno)
continue;
maj = (selecttab[s].filps[fd]->filp_vno->v_sdev >> MAJOR)&BYTE;
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
if(dmap_driver_match(proc_e, maj)) {
select_return(&selecttab[s], EAGAIN);
}
}
}
return;
}