minix/servers/vfs/pipe.c

573 lines
16 KiB
C
Raw Normal View History

2005-04-21 16:53:53 +02:00
/* This file deals with the suspension and revival of processes. A process can
* be suspended because it wants to read or write from a pipe and can't, or
* because it wants to read or write from a special file and can't. When a
* process can't continue it is suspended, and revived later when it is able
* to continue.
*
* The entry points into this file are
* do_pipe2: perform the PIPE2 system call
2005-04-21 16:53:53 +02:00
* pipe_check: check to see that a read or write on a pipe is feasible now
* suspend: suspend a process that cannot do a requested read or write
2005-08-29 18:47:18 +02:00
* release: check to see if a suspended process can be released and do
* it
2005-04-21 16:53:53 +02:00
* revive: mark a suspended process as able to run again
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
* unsuspend_by_endpt: revive all processes blocking on a given process
2005-04-21 16:53:53 +02:00
* do_unpause: a signal has been sent to a process; see if it suspended
*/
#include "fs.h"
#include <fcntl.h>
#include <signal.h>
#include <string.h>
#include <assert.h>
2005-04-21 16:53:53 +02:00
#include <minix/callnr.h>
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
#include <minix/endpoint.h>
2005-04-21 16:53:53 +02:00
#include <minix/com.h>
#include <minix/u64.h>
#include <sys/select.h>
#include <sys/time.h>
2005-04-21 16:53:53 +02:00
#include "file.h"
2012-02-13 16:28:04 +01:00
#include "scratchpad.h"
#include <minix/vfsif.h>
#include "vnode.h"
#include "vmnt.h"
static int create_pipe(int fil_des[2], int flags);
/*===========================================================================*
* do_pipe2 *
*===========================================================================*/
int do_pipe2(void)
{
/* Perform the pipe2(fil_des[2], flags) system call. */
int r, flags;
int fil_des[2]; /* reply goes here */
flags = job_m_in.m_lc_vfs_pipe2.flags;
r = create_pipe(fil_des, flags);
if (r == OK) {
job_m_out.m_lc_vfs_pipe2.fd0 = fil_des[0];
job_m_out.m_lc_vfs_pipe2.fd1 = fil_des[1];
}
return r;
}
/*===========================================================================*
* create_pipe *
*===========================================================================*/
static int create_pipe(int fil_des[2], int flags)
{
2005-04-21 16:53:53 +02:00
register struct fproc *rfp;
int r;
struct filp *fil_ptr0, *fil_ptr1;
struct vnode *vp;
2012-02-13 16:28:04 +01:00
struct vmnt *vmp;
struct node_details res;
2012-02-13 16:28:04 +01:00
/* Get a lock on PFS */
if ((vmp = find_vmnt(PFS_PROC_NR)) == NULL) panic("PFS gone");
VFS: fix locking bugs .sync and fsync used unnecessarily restrictive locking type .fsync violated locking order by obtaining a vmnt lock after a filp lock .fsync contained a TOCTOU bug .new_node violated locking rules (didn't upgrade lock upon file creation) .do_pipe used unnecessarily restrictive locking type .always lock pipes exclusively; even a read operation might require to do a write on a vnode object (update pipe size) .when opening a file with O_TRUNC, upgrade vnode lock when truncating .utime used unnecessarily restrictive locking type .path parsing: .always acquire VMNT_WRITE or VMNT_EXCL on vmnt and downgrade to VMNT_READ if that was what was actually requested. This prevents the following deadlock scenario: thread A: lock_vmnt(vmp, TLL_READSER); lock_vnode(vp, TLL_READSER); upgrade_vmnt_lock(vmp, TLL_WRITE); thread B: lock_vmnt(vmp, TLL_READ); lock_vnode(vp, TLL_READSER); thread A will be stuck in upgrade_vmnt_lock and thread B is stuck in lock_vnode. This happens when, for example, thread A tries create a new node (open.c:new_node) and thread B tries to do eat_path to change dir (stadir.c:do_chdir). When the path is being resolved, a vnode is always locked with VNODE_OPCL (TLL_READSER) and then downgraded to VNODE_READ if read-only is actually requested. Thread A locks the vmnt with VMNT_WRITE (TLL_READSER) which still allows VMNT_READ locks. Thread B can't acquire a lock on the vnode because thread A has it; Thread A can't upgrade its vmnt lock to VMNT_WRITE (TLL_WRITE) because thread B has a VMNT_READ lock on it. By serializing vmnt locks during path parsing, thread B can only acquire a lock on vmp when thread A has completely finished its operation.
2012-11-30 13:49:53 +01:00
if ((r = lock_vmnt(vmp, VMNT_READ)) != OK) return(r);
2012-02-13 16:28:04 +01:00
/* See if a free vnode is available */
2012-04-23 15:45:14 +02:00
if ((vp = get_free_vnode()) == NULL) {
unlock_vmnt(vmp);
return(err_code);
}
2012-02-13 16:28:04 +01:00
lock_vnode(vp, VNODE_OPCL);
2005-04-21 16:53:53 +02:00
/* Acquire two file descriptors. */
rfp = fp;
if ((r = get_fd(fp, 0, R_BIT, &fil_des[0], &fil_ptr0)) != OK) {
2012-02-13 16:28:04 +01:00
unlock_vnode(vp);
unlock_vmnt(vmp);
return(r);
}
2005-04-21 16:53:53 +02:00
rfp->fp_filp[fil_des[0]] = fil_ptr0;
2012-02-13 16:28:04 +01:00
fil_ptr0->filp_count = 1; /* mark filp in use */
if ((r = get_fd(fp, 0, W_BIT, &fil_des[1], &fil_ptr1)) != OK) {
rfp->fp_filp[fil_des[0]] = NULL;
2012-02-13 16:28:04 +01:00
fil_ptr0->filp_count = 0; /* mark filp free */
unlock_filp(fil_ptr0);
unlock_vnode(vp);
unlock_vmnt(vmp);
return(r);
2005-04-21 16:53:53 +02:00
}
rfp->fp_filp[fil_des[1]] = fil_ptr1;
fil_ptr1->filp_count = 1;
/* Create a named pipe inode on PipeFS */
r = req_newnode(PFS_PROC_NR, fp->fp_effuid, fp->fp_effgid, I_NAMED_PIPE,
NO_DEV, &res);
if (r != OK) {
rfp->fp_filp[fil_des[0]] = NULL;
fil_ptr0->filp_count = 0;
rfp->fp_filp[fil_des[1]] = NULL;
fil_ptr1->filp_count = 0;
2012-02-13 16:28:04 +01:00
unlock_filp(fil_ptr1);
unlock_filp(fil_ptr0);
unlock_vnode(vp);
unlock_vmnt(vmp);
return(r);
}
/* Fill in vnode */
vp->v_fs_e = res.fs_e;
vp->v_mapfs_e = res.fs_e;
2012-02-13 16:28:04 +01:00
vp->v_inode_nr = res.inode_nr;
vp->v_mapinode_nr = res.inode_nr;
vp->v_mode = res.fmode;
2007-08-07 14:52:47 +02:00
vp->v_fs_count = 1;
vp->v_mapfs_count = 1;
2007-08-07 14:52:47 +02:00
vp->v_ref_count = 1;
vp->v_size = 0;
2012-02-13 16:28:04 +01:00
vp->v_vmnt = NULL;
vp->v_dev = NO_DEV;
2005-04-21 16:53:53 +02:00
/* Fill in filp objects */
fil_ptr0->filp_vno = vp;
2007-08-07 14:52:47 +02:00
dup_vnode(vp);
fil_ptr1->filp_vno = vp;
fil_ptr0->filp_flags = O_RDONLY | (flags & ~O_ACCMODE);
fil_ptr1->filp_flags = O_WRONLY | (flags & ~O_ACCMODE);
if (flags & O_CLOEXEC) {
FD_SET(fil_des[0], &rfp->fp_cloexec_set);
FD_SET(fil_des[1], &rfp->fp_cloexec_set);
}
2012-02-13 16:28:04 +01:00
unlock_filps(fil_ptr0, fil_ptr1);
unlock_vmnt(vmp);
2005-04-21 16:53:53 +02:00
return(OK);
}
/*===========================================================================*
* map_vnode *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int map_vnode(vp, map_to_fs_e)
struct vnode *vp;
2012-02-13 16:28:04 +01:00
endpoint_t map_to_fs_e;
{
int r;
2012-02-13 16:28:04 +01:00
struct vmnt *vmp;
struct node_details res;
2012-02-13 16:28:04 +01:00
if(vp->v_mapfs_e != NONE) return(OK); /* Already mapped; nothing to do. */
if ((vmp = find_vmnt(map_to_fs_e)) == NULL)
panic("Can't map to unknown endpoint");
if ((r = lock_vmnt(vmp, VMNT_WRITE)) != OK) {
if (r == EBUSY)
vmp = NULL; /* Already locked, do not unlock */
else
return(r);
}
2012-02-13 16:28:04 +01:00
/* Create a temporary mapping of this inode to another FS. Read and write
* operations on data will be handled by that FS. The rest by the 'original'
* FS that holds the inode. */
2012-02-13 16:28:04 +01:00
if ((r = req_newnode(map_to_fs_e, fp->fp_effuid, fp->fp_effgid, I_NAMED_PIPE,
vp->v_dev, &res)) == OK) {
vp->v_mapfs_e = res.fs_e;
2012-02-13 16:28:04 +01:00
vp->v_mapinode_nr = res.inode_nr;
vp->v_mapfs_count = 1;
}
2012-02-13 16:28:04 +01:00
if (vmp) unlock_vmnt(vmp);
return(r);
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* pipe_check *
*===========================================================================*/
int pipe_check(
struct filp *filp, /* the filp of the pipe */
int rw_flag, /* READING or WRITING */
int oflags, /* flags set by open or fcntl */
int bytes, /* bytes to be read or written (all chunks) */
int notouch /* check only */
)
2005-04-21 16:53:53 +02:00
{
/* Pipes are a little different. If a process reads from an empty pipe for
* which a writer still exists, suspend the reader. If the pipe is empty
* and there is no writer, return 0 bytes. If a process is writing to a
* pipe and no one is reading from it, give a broken pipe error.
*/
struct vnode *vp;
off_t pos;
int r = OK;
vp = filp->filp_vno;
/* Reads start at the beginning; writes append to pipes */
if (notouch) /* In this case we don't actually care whether data transfer
* would succeed. See POSIX 1003.1-2008 */
pos = 0;
else if (rw_flag == READING)
pos = 0;
else {
pos = vp->v_size;
}
2005-04-21 16:53:53 +02:00
/* If reading, check for empty pipe. */
if (rw_flag == READING) {
if (vp->v_size == 0) {
2005-04-21 16:53:53 +02:00
/* Process is reading from an empty pipe. */
if (find_filp(vp, W_BIT) != NULL) {
2005-04-21 16:53:53 +02:00
/* Writer exists */
2012-02-13 16:28:04 +01:00
if (oflags & O_NONBLOCK)
2005-04-21 16:53:53 +02:00
r = EAGAIN;
2012-02-13 16:28:04 +01:00
else
2005-04-21 16:53:53 +02:00
r = SUSPEND;
2012-02-13 16:28:04 +01:00
2005-04-21 16:53:53 +02:00
/* If need be, activate sleeping writers. */
/* We ignore notouch voluntary here. */
if (susp_count > 0)
release(vp, VFS_WRITE, susp_count);
2005-04-21 16:53:53 +02:00
}
return(r);
}
return(bytes);
2007-08-07 14:52:47 +02:00
}
/* Process is writing to a pipe. */
if (find_filp(vp, R_BIT) == NULL) {
2007-08-07 14:52:47 +02:00
return(EPIPE);
}
2005-04-21 16:53:53 +02:00
/* Calculate how many bytes can be written. */
2007-08-07 14:52:47 +02:00
if (pos + bytes > PIPE_BUF) {
if (oflags & O_NONBLOCK) {
2007-08-07 14:52:47 +02:00
if (bytes <= PIPE_BUF) {
/* Write has to be atomic */
2005-04-21 16:53:53 +02:00
return(EAGAIN);
}
2007-08-07 14:52:47 +02:00
/* Compute available space */
bytes = PIPE_BUF - pos;
2007-08-07 14:52:47 +02:00
if (bytes > 0) {
/* Do a partial write. Need to wakeup reader */
if (!notouch)
release(vp, VFS_READ, susp_count);
2007-08-07 14:52:47 +02:00
return(bytes);
} else {
/* Pipe is full */
return(EAGAIN);
}
}
if (bytes > PIPE_BUF) {
/* Compute available space */
bytes = PIPE_BUF - pos;
2007-08-07 14:52:47 +02:00
if (bytes > 0) {
/* Do a partial write. Need to wakeup reader
* since we'll suspend ourself in read_write()
*/
if (!notouch)
release(vp, VFS_READ, susp_count);
2007-08-07 14:52:47 +02:00
return(bytes);
2005-04-21 16:53:53 +02:00
}
}
/* Pipe is full */
2007-08-07 14:52:47 +02:00
return(SUSPEND);
2005-04-21 16:53:53 +02:00
}
2007-08-07 14:52:47 +02:00
/* Writing to an empty pipe. Search for suspended reader. */
if (pos == 0 && !notouch)
release(vp, VFS_READ, susp_count);
2007-08-07 14:52:47 +02:00
/* Requested amount fits */
return(bytes);
2005-04-21 16:53:53 +02:00
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* suspend *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
void suspend(int why)
2005-04-21 16:53:53 +02:00
{
/* Take measures to suspend the processing of the present system call.
* Store the parameters to be used upon resuming in the process table.
* (Actually they are not used when a process is waiting for an I/O device,
* but they are needed for pipes, and it is not worth making the distinction.)
* The SUSPEND pseudo error should be returned after calling suspend().
*/
2012-02-13 16:28:04 +01:00
if (why == FP_BLOCKED_ON_POPEN || why == FP_BLOCKED_ON_PIPE)
/* #procs susp'ed on pipe*/
susp_count++;
fp->fp_blocked_on = why;
assert(fp->fp_grant == GRANT_INVALID || !GRANT_VALID(fp->fp_grant));
fp->fp_block_callnr = job_call_nr;
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* wait_for *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
void wait_for(endpoint_t who)
{
if(who == NONE || who == ANY)
panic("suspend on NONE or ANY");
suspend(FP_BLOCKED_ON_OTHER);
fp->fp_task = who;
}
2007-08-07 14:52:47 +02:00
/*===========================================================================*
* pipe_suspend *
2007-08-07 14:52:47 +02:00
*===========================================================================*/
2012-03-25 20:25:53 +02:00
void pipe_suspend(filp, buf, size)
2012-02-13 16:28:04 +01:00
struct filp *filp;
vir_bytes buf;
2007-08-07 14:52:47 +02:00
size_t size;
{
/* Take measures to suspend the processing of the present system call.
* Store the parameters to be used upon resuming in the process table.
*/
2012-02-13 16:28:04 +01:00
scratch(fp).file.filp = filp;
scratch(fp).io.io_buffer = buf;
scratch(fp).io.io_nbytes = size;
suspend(FP_BLOCKED_ON_PIPE);
2007-08-07 14:52:47 +02:00
}
/*===========================================================================*
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
* unsuspend_by_endpt *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
void unsuspend_by_endpt(endpoint_t proc_e)
{
/* Revive processes waiting for drivers (SUSPENDed) that have disappeared with
* return code EAGAIN.
*/
struct fproc *rp;
for (rp = &fproc[0]; rp < &fproc[NR_PROCS]; rp++) {
2012-02-13 16:28:04 +01:00
if (rp->fp_pid == PID_FREE) continue;
if (rp->fp_blocked_on == FP_BLOCKED_ON_OTHER && rp->fp_task == proc_e)
revive(rp->fp_endpoint, EIO);
}
/* Revive processes waiting in drivers on select()s with EAGAIN too */
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(proc_e);
return;
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* release *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
void release(vp, op, count)
register struct vnode *vp; /* inode of pipe */
int op; /* VFS_READ, VFS_WRITE, or VFS_OPEN */
2005-04-21 16:53:53 +02:00
int count; /* max number of processes to release */
{
2012-02-13 16:28:04 +01:00
/* Check to see if any process is hanging on vnode 'vp'. If one is, and it
* was trying to perform the call indicated by 'op', release it.
2005-04-21 16:53:53 +02:00
*/
register struct fproc *rp;
struct filp *f;
2012-02-13 16:28:04 +01:00
int selop;
/* Trying to perform the call also includes SELECTing on it with that
* operation.
*/
if (op == VFS_READ || op == VFS_WRITE) {
if (op == VFS_READ)
2012-02-13 16:28:04 +01:00
selop = SEL_RD;
else
selop = SEL_WR;
for (f = &filp[0]; f < &filp[NR_FILPS]; f++) {
if (f->filp_count < 1 || !(f->filp_pipe_select_ops & selop) ||
f->filp_vno != vp)
continue;
2012-02-13 16:28:04 +01:00
select_callback(f, selop);
2012-02-13 16:28:04 +01:00
f->filp_pipe_select_ops &= ~selop;
}
}
2005-04-21 16:53:53 +02:00
/* Search the proc table. */
2009-04-02 13:44:26 +02:00
for (rp = &fproc[0]; rp < &fproc[NR_PROCS] && count > 0; rp++) {
if (rp->fp_pid != PID_FREE && fp_is_blocked(rp) &&
2012-02-13 16:28:04 +01:00
!(rp->fp_flags & FP_REVIVED) && rp->fp_block_callnr == op) {
/* Find the vnode. Depending on the reason the process was
* suspended, there are different ways of finding it.
*/
if (rp->fp_blocked_on == FP_BLOCKED_ON_POPEN ||
rp->fp_blocked_on == FP_BLOCKED_ON_LOCK ||
rp->fp_blocked_on == FP_BLOCKED_ON_OTHER) {
f = rp->fp_filp[scratch(rp).file.fd_nr];
if (f == NULL || f->filp_mode == FILP_CLOSED)
2012-02-13 16:28:04 +01:00
continue;
if (rp->fp_filp[scratch(rp).file.fd_nr]->filp_vno != vp)
continue;
} else if (rp->fp_blocked_on == FP_BLOCKED_ON_PIPE) {
if (scratch(rp).file.filp == NULL)
continue;
if (scratch(rp).file.filp->filp_vno != vp)
continue;
} else
continue;
/* We found the vnode. Revive process. */
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(rp->fp_endpoint, 0);
2005-04-21 16:53:53 +02:00
susp_count--; /* keep track of who is suspended */
2009-04-02 13:44:26 +02:00
if(susp_count < 0)
panic("susp_count now negative: %d", susp_count);
2005-04-21 16:53:53 +02:00
if (--count == 0) return;
}
}
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* revive *
*===========================================================================*/
void revive(endpoint_t proc_e, int returned)
2005-04-21 16:53:53 +02:00
{
/* Revive a previously blocked process. When a process hangs on tty, this
* is the way it is eventually released. For processes blocked on _SELECT and
* _OTHER, this function MUST NOT block its calling thread.
2005-04-21 16:53:53 +02:00
*/
struct fproc *rfp;
int blocked_on;
2012-02-13 16:28:04 +01:00
int fd_nr, slot;
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 (proc_e == NONE || isokendpt(proc_e, &slot) != OK) return;
2005-04-21 16:53:53 +02:00
2012-02-13 16:28:04 +01:00
rfp = &fproc[slot];
if (!fp_is_blocked(rfp) || (rfp->fp_flags & FP_REVIVED)) return;
2005-04-21 16:53:53 +02:00
/* The 'reviving' flag only applies to pipes. Processes waiting for TTY get
* a message right away. The revival process is different for TTY and pipes.
* For select and TTY revival, the work is already done, for pipes it is not:
2012-02-13 16:28:04 +01:00
* the proc must be restarted so it can try again.
2005-04-21 16:53:53 +02:00
*/
blocked_on = rfp->fp_blocked_on;
2012-02-13 16:28:04 +01:00
fd_nr = scratch(rfp).file.fd_nr;
if (blocked_on == FP_BLOCKED_ON_PIPE || blocked_on == FP_BLOCKED_ON_LOCK) {
2005-04-21 16:53:53 +02:00
/* Revive a process suspended on a pipe or lock. */
2012-02-13 16:28:04 +01:00
rfp->fp_flags |= FP_REVIVED;
2005-04-21 16:53:53 +02:00
reviving++; /* process was waiting on pipe or lock */
} else {
rfp->fp_blocked_on = FP_BLOCKED_ON_NONE;
2012-02-13 16:28:04 +01:00
scratch(rfp).file.fd_nr = 0;
if (blocked_on == FP_BLOCKED_ON_POPEN) {
/* process blocked in open or create */
replycode(proc_e, fd_nr);
} else if (blocked_on == FP_BLOCKED_ON_SELECT) {
replycode(proc_e, returned);
} else {
2012-02-13 16:28:04 +01:00
/* Revive a process suspended on TTY or other device.
* Pretend it wants only what there is.
*/
2012-02-13 16:28:04 +01:00
scratch(rfp).io.io_nbytes = returned;
/* If a grant has been issued by FS for this I/O, revoke
* it again now that I/O is done.
*/
2012-02-13 16:28:04 +01:00
if (GRANT_VALID(rfp->fp_grant)) {
if(cpf_revoke(rfp->fp_grant)) {
2012-02-13 16:28:04 +01:00
panic("VFS: revoke failed for grant: %d",
rfp->fp_grant);
2012-02-13 16:28:04 +01:00
}
rfp->fp_grant = GRANT_INVALID;
}
replycode(proc_e, returned);/* unblock the process */
2005-04-21 16:53:53 +02:00
}
}
}
2006-05-11 16:57:23 +02:00
/*===========================================================================*
* unpause *
*===========================================================================*/
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
void unpause(void)
2006-05-11 16:57:23 +02:00
{
2005-04-21 16:53:53 +02:00
/* A signal has been sent to a user who is paused on the file system.
* Abort the system call with the EINTR error message.
*/
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
int blocked_on, fild, status = EINTR;
2005-04-21 16:53:53 +02:00
struct filp *f;
dev_t dev;
int wasreviving = 0;
2005-04-21 16:53:53 +02:00
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
if (!fp_is_blocked(fp)) return;
blocked_on = fp->fp_blocked_on;
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
/* Clear the block status now. The procedure below might make blocking calls
* and it is imperative that while at least cdev_cancel() is executing, other
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
* parts of VFS do not perceive this process as blocked on something.
*/
fp->fp_blocked_on = FP_BLOCKED_ON_NONE;
2005-04-21 16:53:53 +02:00
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
if (fp->fp_flags & FP_REVIVED) {
fp->fp_flags &= ~FP_REVIVED;
reviving--;
wasreviving = 1;
}
switch (blocked_on) {
case FP_BLOCKED_ON_PIPE:/* process trying to read or write a pipe */
/* If the operation succeeded partially, return the bytes
* processed so far, and clear the remembered state. Otherwise,
* return EINTR as usual.
*/
if (fp->fp_cum_io_partial > 0) {
status = fp->fp_cum_io_partial;
fp->fp_cum_io_partial = 0;
}
2005-04-21 16:53:53 +02:00
break;
case FP_BLOCKED_ON_LOCK:/* process trying to set a lock with FCNTL */
2005-04-21 16:53:53 +02:00
break;
case FP_BLOCKED_ON_SELECT:/* process blocking on select() */
select_forget();
break;
2012-02-13 16:28:04 +01:00
case FP_BLOCKED_ON_POPEN: /* process trying to open a fifo */
2005-04-21 16:53:53 +02:00
break;
2012-02-13 16:28:04 +01:00
case FP_BLOCKED_ON_OTHER:/* process trying to do device I/O (e.g. tty)*/
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
fild = scratch(fp).file.fd_nr;
if (fild < 0 || fild >= OPEN_MAX)
2012-02-13 16:28:04 +01:00
panic("file descriptor out-of-range");
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
f = fp->fp_filp[fild];
if(!f) {
sys_diagctl_stacktrace(fp->fp_endpoint);
panic("process %d blocked on empty fd %d",
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
fp->fp_endpoint, fild);
}
dev = f->filp_vno->v_sdev; /* device hung on */
2012-02-13 16:28:04 +01:00
status = cdev_cancel(dev);
break;
default :
2012-02-13 16:28:04 +01:00
panic("VFS: unknown block reason: %d", blocked_on);
2005-04-21 16:53:53 +02:00
}
2012-02-13 16:28:04 +01:00
if ((blocked_on == FP_BLOCKED_ON_PIPE || blocked_on == FP_BLOCKED_ON_POPEN)&&
!wasreviving) {
susp_count--;
}
VFS: worker thread model overhaul The main purpose of this patch is to fix handling of unpause calls from PM while another call is ongoing. The solution to this problem sparked a full revision of the threading model, consisting of a large number of related changes: - all active worker threads are now always associated with a process, and every process has at most one active thread working for it; - the process lock is always held by a process's worker thread; - a process can now have both normal work and postponed PM work associated to it; - timer expiry and non-postponed PM work is done from the main thread; - filp garbage collection is done from a thread associated with VFS; - reboot calls from PM are now done from a thread associated with PM; - the DS events handler is protected from starting multiple threads; - support for a system worker thread has been removed; - the deadlock recovery thread has been replaced by a parameter to the worker_start() function; the number of worker threads has consequently been increased by one; - saving and restoring of global but per-thread variables is now centralized in worker_suspend() and worker_resume(); err_code is now saved and restored in all cases; - the concept of jobs has been removed, and job_m_in now points to a message stored in the worker thread structure instead; - the PM lock has been removed; - the separate exec lock has been replaced by a lock on the VM process, which was already being locked for exec calls anyway; - PM_UNPAUSE is now processed as a postponed PM request, from a thread associated with the target process; - the FP_DROP_WORK flag has been removed, since it is no longer more than just an optimization and only applied to processes operating on a pipe when getting killed; - assignment to "fp" now takes place only when obtaining new work in the main thread or a worker thread, when resuming execution of a thread, and in the special case of exiting processes during reboot; - there are no longer special cases where the yield() call is used to force a thread to run. Change-Id: I7a97b9b95c2450454a9b5318dfa0e6150d4e6858
2013-08-30 14:00:50 +02:00
replycode(fp->fp_endpoint, status); /* signal interrupted call */
}