toaruos/kernel/sys/process.c

980 lines
26 KiB
C

/* vim: tabstop=4 shiftwidth=4 noexpandtab
* This file is part of ToaruOS and is released under the terms
* of the NCSA / University of Illinois License - see LICENSE.md
* Copyright (C) 2011-2018 K. Lange
* Copyright (C) 2012 Markus Schober
* Copyright (C) 2015 Dale Weiler
*
* Processes
*
* Internal format format for a process and functions to spawn
* new processes and manage the process tree.
*/
#include <kernel/system.h>
#include <kernel/process.h>
#include <kernel/bitset.h>
#include <kernel/logging.h>
#include <kernel/shm.h>
#include <kernel/printf.h>
#include <sys/wait.h>
#include <toaru/list.h>
#include <toaru/tree.h>
tree_t * process_tree; /* Parent->Children tree */
list_t * process_list; /* Flat storage */
list_t * process_queue; /* Ready queue */
list_t * sleep_queue;
volatile process_t * current_process = NULL;
process_t * kernel_idle_task = NULL;
static spin_lock_t tree_lock = { 0 };
static spin_lock_t process_queue_lock = { 0 };
static spin_lock_t wait_lock_tmp = { 0 };
static spin_lock_t sleep_lock = { 0 };
static bitset_t pid_set;
/* Default process name string */
char * default_name = "[unnamed]";
int is_valid_process(process_t * process) {
foreach(lnode, process_list) {
if (lnode->value == process) {
return 1;
}
}
return 0;
}
/*
* This makes a nice 4096-byte bitmap. It also happens
* to be pid_max on 32-bit Linux, so that's kinda nice.
*/
#define MAX_PID 32768
/*
* Initialize the process tree and ready queue.
*/
void initialize_process_tree(void) {
process_tree = tree_create();
process_list = list_create();
process_queue = list_create();
sleep_queue = list_create();
/* Start off with enough bits for 64 processes */
bitset_init(&pid_set, MAX_PID / 8);
/* First two bits are set by default */
bitset_set(&pid_set, 0);
bitset_set(&pid_set, 1);
}
/*
* Recursively print a process node to the console.
*
* @param node Node to print.
* @param height Current depth in the tree.
*/
void debug_print_process_tree_node(tree_node_t * node, size_t height) {
/* End recursion on a blank entry */
if (!node) return;
char * tmp = malloc(512);
memset(tmp, 0, 512);
char * c = tmp;
/* Indent output */
for (uint32_t i = 0; i < height; ++i) {
c += sprintf(c, " ");
}
/* Get the current process */
process_t * proc = (process_t *)node->value;
/* Print the process name */
c += sprintf(c, "%d.%d %s", proc->group ? proc->group : proc->id, proc->id, proc->name);
if (proc->description) {
/* And, if it has one, its description */
c += sprintf(c, " %s", proc->description);
}
if (proc->finished) {
c += sprintf(c, " [zombie]");
}
/* Linefeed */
debug_print(NOTICE, "%s", tmp);
free(tmp);
foreach(child, node->children) {
/* Recursively print the children */
debug_print_process_tree_node(child->value, height + 1);
}
}
/*
* Print the process tree to the console.
*/
void debug_print_process_tree(void) {
debug_print_process_tree_node(process_tree->root, 0);
}
/*
* Retreive the next ready process.
* XXX: POPs from the ready queue!
*
* @return A pointer to the next process in the queue.
*/
process_t * next_ready_process(void) {
if (!process_available()) {
return kernel_idle_task;
}
if (process_queue->head->owner != process_queue) {
debug_print(ERROR, "Erroneous process located in process queue: node 0x%x has owner 0x%x, but process_queue is 0x%x", process_queue->head, process_queue->head->owner, process_queue);
process_t * proc = process_queue->head->value;
debug_print(ERROR, "PID associated with this node is %d", proc->id);
}
node_t * np = list_dequeue(process_queue);
assert(np && "Ready queue is empty.");
process_t * next = np->value;
return next;
}
/*
* Reinsert a process into the ready queue.
*
* @param proc Process to reinsert
*/
void make_process_ready(process_t * proc) {
if (proc->sleep_node.owner != NULL) {
if (proc->sleep_node.owner == sleep_queue) {
/* XXX can't wake from timed sleep */
if (proc->timed_sleep_node) {
IRQ_OFF;
spin_lock(sleep_lock);
list_delete(sleep_queue, proc->timed_sleep_node);
spin_unlock(sleep_lock);
IRQ_RES;
proc->sleep_node.owner = NULL;
free(proc->timed_sleep_node->value);
}
/* Else: I have no idea what happened. */
} else {
proc->sleep_interrupted = 1;
spin_lock(wait_lock_tmp);
list_delete((list_t*)proc->sleep_node.owner, &proc->sleep_node);
spin_unlock(wait_lock_tmp);
}
}
if (proc->sched_node.owner) {
debug_print(WARNING, "Can't make process ready without removing from owner list: %d", proc->id);
debug_print(WARNING, " (This is a bug) Current owner list is 0x%x (ready queue is 0x%x)", proc->sched_node.owner, process_queue);
return;
}
spin_lock(process_queue_lock);
list_append(process_queue, &proc->sched_node);
spin_unlock(process_queue_lock);
}
extern void tree_remove_reparent_root(tree_t * tree, tree_node_t * node);
/*
* Delete a process from the process tree
*
* @param proc Process to find and remove.
*/
void delete_process(process_t * proc) {
tree_node_t * entry = proc->tree_entry;
/* The process must exist in the tree, or the client is at fault */
if (!entry) return;
/* We can not remove the root, which is an error anyway */
assert((entry != process_tree->root) && "Attempted to kill init.");
if (process_tree->root == entry) {
/* We are init, don't even bother. */
return;
}
/* Remove the entry. */
spin_lock(tree_lock);
/* Reparent everyone below me to init */
int has_children = entry->children->length;
tree_remove_reparent_root(process_tree, entry);
list_delete(process_list, list_find(process_list, proc));
spin_unlock(tree_lock);
if (has_children) {
process_t * init = process_tree->root->value;
wakeup_queue(init->wait_queue);
}
bitset_clear(&pid_set, proc->id);
/* Uh... */
free(proc);
}
static void _kidle(void) {
while (1) {
IRQ_ON;
PAUSE;
}
}
/*
* Spawn the idle "process".
*/
process_t * spawn_kidle(void) {
process_t * idle = malloc(sizeof(process_t));
memset(idle, 0x00, sizeof(process_t));
idle->id = -1;
idle->name = strdup("[kidle]");
idle->is_tasklet = 1;
idle->image.stack = (uintptr_t)malloc(KERNEL_STACK_SIZE) + KERNEL_STACK_SIZE;
idle->thread.eip = (uintptr_t)&_kidle;
idle->thread.esp = idle->image.stack;
idle->thread.ebp = idle->image.stack;
idle->started = 1;
idle->running = 1;
idle->wait_queue = list_create();
idle->shm_mappings = list_create();
idle->signal_queue = list_create();
gettimeofday(&idle->start, NULL);
set_process_environment(idle, current_directory);
return idle;
}
/*
* Spawn the initial process.
*
* @return A pointer to the new initial process entry
*/
process_t * spawn_init(void) {
/* We can only do this once. */
assert((!process_tree->root) && "Tried to regenerate init!");
/* Allocate space for a new process */
process_t * init = malloc(sizeof(process_t));
/* Set it as the root process */
tree_set_root(process_tree, (void *)init);
/* Set its tree entry pointer so we can keep track
* of the process' entry in the process tree. */
init->tree_entry = process_tree->root;
init->id = 1; /* Init is PID 1 */
init->group = 0; /* thread group id (real PID) */
init->job = 1; /* process group id (jobs) */
init->session = 1; /* session leader id */
init->name = strdup("init"); /* Um, duh. */
init->cmdline = NULL;
init->user = 0; /* UID 0 */
init->mask = 022; /* umask */
init->status = 0; /* Run status */
init->fds = malloc(sizeof(fd_table_t));
init->fds->refs = 1;
init->fds->length = 0; /* Initialize the file descriptors */
init->fds->capacity = 4;
init->fds->entries = malloc(sizeof(fs_node_t *) * init->fds->capacity);
/* Set the working directory */
init->wd_node = clone_fs(fs_root);
init->wd_name = strdup("/");
/* Heap and stack pointers (and actuals) */
init->image.entry = 0;
init->image.heap = 0;
init->image.heap_actual = 0;
init->image.stack = initial_esp + 1;
init->image.user_stack = 0;
init->image.size = 0;
init->image.shm_heap = SHM_START; /* Yeah, a bit of a hack. */
spin_init(init->image.lock);
/* Process is not finished */
init->finished = 0;
init->suspended = 0;
init->started = 1;
init->running = 1;
init->wait_queue = list_create();
init->shm_mappings = list_create();
init->signal_queue = list_create();
init->signal_kstack = NULL; /* None yet initialized */
init->sched_node.prev = NULL;
init->sched_node.next = NULL;
init->sched_node.value = init;
init->sleep_node.prev = NULL;
init->sleep_node.next = NULL;
init->sleep_node.value = init;
init->timed_sleep_node = NULL;
init->is_tasklet = 0;
set_process_environment(init, current_directory);
/* What the hey, let's also set the description on this one */
init->description = strdup("[init]");
list_insert(process_list, (void *)init);
return init;
}
/*
* Get the next available PID
*
* @return A usable PID for a new process.
*/
static int _next_pid = 2;
pid_t get_next_pid(void) {
if (_next_pid > MAX_PID) {
int index = bitset_ffub(&pid_set);
/*
* Honestly, we don't have the memory to really risk reaching
* the point where we have MAX_PID processes running
* concurrently, so this assertion should be "safe enough".
*/
assert(index != -1);
bitset_set(&pid_set, index);
return index;
}
int pid = _next_pid;
_next_pid++;
assert(!bitset_test(&pid_set, pid) && "Next PID already allocated?");
bitset_set(&pid_set, pid);
return pid;
}
/*
* Disown a process from its parent.
*/
void process_disown(process_t * proc) {
assert(process_tree->root && "No init, has the process tree been initialized?");
/* Find the process in the tree */
tree_node_t * entry = proc->tree_entry;
/* Break it of from its current parent */
spin_lock(tree_lock);
tree_break_off(process_tree, entry);
/* And insert it back elsewhere */
tree_node_insert_child_node(process_tree, process_tree->root, entry);
spin_unlock(tree_lock);
}
/*
* Spawn a new process.
*
* @param parent The parent process to spawn the new one off of.
* @return A pointer to the new process.
*/
process_t * spawn_process(volatile process_t * parent, int reuse_fds) {
assert(process_tree->root && "Attempted to spawn a process without init.");
/* Allocate a new process */
debug_print(INFO," process_t {");
process_t * proc = calloc(sizeof(process_t),1);
debug_print(INFO," }");
proc->id = get_next_pid(); /* Set its PID */
proc->group = proc->id; /* Set the GID */
proc->name = strdup(parent->name); /* Use the default name */
proc->description = NULL; /* No description */
proc->cmdline = parent->cmdline;
/* Copy permissions */
proc->user = parent->user;
proc->mask = parent->mask;
/* Until specified otherwise */
proc->job = parent->job;
proc->session = parent->session;
/* Zero out the ESP/EBP/EIP */
proc->thread.esp = 0;
proc->thread.ebp = 0;
proc->thread.eip = 0;
proc->thread.fpu_enabled = 0;
memcpy((void*)proc->thread.fp_regs, (void*)parent->thread.fp_regs, 512);
/* Set the process image information from the parent */
proc->image.entry = parent->image.entry;
proc->image.heap = parent->image.heap;
proc->image.heap_actual = parent->image.heap_actual;
proc->image.size = parent->image.size;
debug_print(INFO," stack {");
proc->image.stack = (uintptr_t)kvmalloc(KERNEL_STACK_SIZE) + KERNEL_STACK_SIZE;
debug_print(INFO," }");
proc->image.user_stack = parent->image.user_stack;
proc->image.shm_heap = SHM_START; /* Yeah, a bit of a hack. */
spin_init(proc->image.lock);
assert(proc->image.stack && "Failed to allocate kernel stack for new process.");
/* Clone the file descriptors from the original process */
if (reuse_fds) {
proc->fds = parent->fds;
proc->fds->refs++;
} else {
proc->fds = malloc(sizeof(fd_table_t));
proc->fds->refs = 1;
proc->fds->length = parent->fds->length;
proc->fds->capacity = parent->fds->capacity;
debug_print(INFO," fds / files {");
proc->fds->entries = malloc(sizeof(fs_node_t *) * proc->fds->capacity);
assert(proc->fds->entries && "Failed to allocate file descriptor table for new process.");
debug_print(INFO," ---");
for (uint32_t i = 0; i < parent->fds->length; ++i) {
proc->fds->entries[i] = clone_fs(parent->fds->entries[i]);
}
debug_print(INFO," }");
}
/* As well as the working directory */
proc->wd_node = clone_fs(parent->wd_node);
proc->wd_name = strdup(parent->wd_name);
/* Zero out the process status */
proc->status = 0;
proc->finished = 0;
proc->suspended = 0;
proc->started = 0;
proc->running = 0;
memset(proc->signals.functions, 0x00, sizeof(uintptr_t) * NUMSIGNALS);
proc->wait_queue = list_create();
proc->shm_mappings = list_create();
proc->signal_queue = list_create();
proc->signal_kstack = NULL; /* None yet initialized */
proc->sched_node.prev = NULL;
proc->sched_node.next = NULL;
proc->sched_node.value = proc;
proc->sleep_node.prev = NULL;
proc->sleep_node.next = NULL;
proc->sleep_node.value = proc;
proc->timed_sleep_node = NULL;
proc->is_tasklet = 0;
gettimeofday(&proc->start, NULL);
/* Insert the process into the process tree as a child
* of the parent process. */
tree_node_t * entry = tree_node_create(proc);
assert(entry && "Failed to allocate a process tree node for new process.");
proc->tree_entry = entry;
spin_lock(tree_lock);
tree_node_insert_child_node(process_tree, parent->tree_entry, entry);
list_insert(process_list, (void *)proc);
spin_unlock(tree_lock);
/* Return the new process */
return proc;
}
uint8_t process_compare(void * proc_v, void * pid_v) {
pid_t pid = (*(pid_t *)pid_v);
process_t * proc = (process_t *)proc_v;
return (uint8_t)(proc->id == pid);
}
process_t * process_from_pid(pid_t pid) {
if (pid < 0) return NULL;
spin_lock(tree_lock);
tree_node_t * entry = tree_find(process_tree,&pid,process_compare);
spin_unlock(tree_lock);
if (entry) {
return (process_t *)entry->value;
}
return NULL;
}
process_t * process_get_parent(process_t * process) {
process_t * result = NULL;
spin_lock(tree_lock);
tree_node_t * entry = process->tree_entry;
if (entry->parent) {
result = entry->parent->value;
}
spin_unlock(tree_lock);
return result;
}
/*
* Wait for children.
*
* @param process Process doing the waiting.
* @param pid PID to wait for
* @param status [out] Where to put the status conditions of the waited-for process
* @param options Options (unused)
* @return A pointer to the process that broke the wait
*/
process_t * process_wait(process_t * process, pid_t pid, int * status, int options) {
/* `options` is ignored */
if (pid == -1) {
/* wait for any child process */
} else if (pid < 0) {
/* wait for any porcess whose ->group == processes[abs(pid)]->group */
} else if (pid == 0) {
/* wait for any process whose ->group == process->group */
} else {
/* wait for processes[pid] */
}
return NULL;
}
/*
* Wake up a sleeping process
*
* @param process Process to wake up
* @param caller Who woke it up
* @return Don't know yet, but I think it should return something.
*/
int process_wake(process_t * process, process_t * caller) {
return 0;
}
/*
* Set the directory for a process.
*
* @param proc Process to set the directory for.
* @param directory Directory to set.
*/
void set_process_environment(process_t * proc, page_directory_t * directory) {
assert(proc);
assert(directory);
proc->thread.page_directory = directory;
}
/*
* Are there any processes available in the queue?
* (Queue not empty)
*
* @return 1 if there are processes available, 0 otherwise
*/
uint8_t process_available(void) {
return (process_queue->head != NULL);
}
/*
* Append a file descriptor to a process.
*
* @param proc Process to append to
* @param node The VFS node
* @return The actual fd, for use in userspace
*/
uint32_t process_append_fd(process_t * proc, fs_node_t * node) {
/* Fill gaps */
for (unsigned int i = 0; i < proc->fds->length; ++i) {
if (!proc->fds->entries[i]) {
proc->fds->entries[i] = node;
return i;
}
}
/* No gaps, expand */
if (proc->fds->length == proc->fds->capacity) {
proc->fds->capacity *= 2;
proc->fds->entries = realloc(proc->fds->entries, sizeof(fs_node_t *) * proc->fds->capacity);
}
proc->fds->entries[proc->fds->length] = node;
proc->fds->length++;
return proc->fds->length-1;
}
/*
* dup2() -> Move the file pointed to by `s(ou)rc(e)` into
* the slot pointed to be `dest(ination)`.
*
* @param proc Process to do this for
* @param src Source file descriptor
* @param dest Destination file descriptor
* @return The destination file descriptor, -1 on failure
*/
uint32_t process_move_fd(process_t * proc, int src, int dest) {
if ((size_t)src > proc->fds->length || (dest != -1 && (size_t)dest > proc->fds->length)) {
return -1;
}
if (dest == -1) {
dest = process_append_fd(proc, NULL);
}
if (proc->fds->entries[dest] != proc->fds->entries[src]) {
close_fs(proc->fds->entries[dest]);
proc->fds->entries[dest] = proc->fds->entries[src];
open_fs(proc->fds->entries[dest], 0);
}
return dest;
}
int wakeup_queue(list_t * queue) {
int awoken_processes = 0;
while (queue->length > 0) {
spin_lock(wait_lock_tmp);
node_t * node = list_pop(queue);
spin_unlock(wait_lock_tmp);
if (!((process_t *)node->value)->finished) {
make_process_ready(node->value);
}
awoken_processes++;
}
return awoken_processes;
}
int wakeup_queue_interrupted(list_t * queue) {
int awoken_processes = 0;
while (queue->length > 0) {
spin_lock(wait_lock_tmp);
node_t * node = list_pop(queue);
spin_unlock(wait_lock_tmp);
if (!((process_t *)node->value)->finished) {
process_t * proc = node->value;
proc->sleep_interrupted = 1;
make_process_ready(proc);
}
awoken_processes++;
}
return awoken_processes;
}
int sleep_on(list_t * queue) {
if (current_process->sleep_node.owner) {
/* uh, we can't sleep right now, we're marked as ready */
switch_task(0);
return 0;
}
current_process->sleep_interrupted = 0;
spin_lock(wait_lock_tmp);
list_append(queue, (node_t *)&current_process->sleep_node);
spin_unlock(wait_lock_tmp);
switch_task(0);
return current_process->sleep_interrupted;
}
int process_is_ready(process_t * proc) {
return (proc->sched_node.owner != NULL);
}
void wakeup_sleepers(unsigned long seconds, unsigned long subseconds) {
IRQ_OFF;
spin_lock(sleep_lock);
if (sleep_queue->length) {
sleeper_t * proc = ((sleeper_t *)sleep_queue->head->value);
while (proc && (proc->end_tick < seconds || (proc->end_tick == seconds && proc->end_subtick <= subseconds))) {
if (proc->is_fswait) {
proc->is_fswait = -1;
process_alert_node(proc->process,proc);
} else {
process_t * process = proc->process;
process->sleep_node.owner = NULL;
process->timed_sleep_node = NULL;
if (!process_is_ready(process)) {
make_process_ready(process);
}
}
free(proc);
free(list_dequeue(sleep_queue));
if (sleep_queue->length) {
proc = ((sleeper_t *)sleep_queue->head->value);
} else {
break;
}
}
}
spin_unlock(sleep_lock);
IRQ_RES;
}
void sleep_until(process_t * process, unsigned long seconds, unsigned long subseconds) {
if (current_process->sleep_node.owner) {
/* Can't sleep, sleeping already */
return;
}
process->sleep_node.owner = sleep_queue;
IRQ_OFF;
spin_lock(sleep_lock);
node_t * before = NULL;
foreach(node, sleep_queue) {
sleeper_t * candidate = ((sleeper_t *)node->value);
if (candidate->end_tick > seconds || (candidate->end_tick == seconds && candidate->end_subtick > subseconds)) {
break;
}
before = node;
}
sleeper_t * proc = malloc(sizeof(sleeper_t));
proc->process = process;
proc->end_tick = seconds;
proc->end_subtick = subseconds;
proc->is_fswait = 0;
process->timed_sleep_node = list_insert_after(sleep_queue, before, proc);
spin_unlock(sleep_lock);
IRQ_RES;
}
void cleanup_process(process_t * proc, int retval) {
proc->status = retval;
proc->finished = 1;
list_free(proc->wait_queue);
free(proc->wait_queue);
list_free(proc->signal_queue);
free(proc->signal_queue);
free(proc->wd_name);
if (proc->node_waits) {
list_free(proc->node_waits);
free(proc->node_waits);
proc->node_waits = NULL;
}
debug_print(INFO, "Releasing shared memory for %d", proc->id);
shm_release_all(proc);
free(proc->shm_mappings);
debug_print(INFO, "Freeing more mems %d", proc->id);
if (proc->signal_kstack) {
free(proc->signal_kstack);
}
release_directory(proc->thread.page_directory);
debug_print(INFO, "Dec'ing fds for %d", proc->id);
proc->fds->refs--;
if (proc->fds->refs == 0) {
debug_print(INFO, "Reached 0, all dependencies are closed for %d's file descriptors and page directories", proc->id);
debug_print(INFO, "Going to clear out the file descriptors %d", proc->id);
for (uint32_t i = 0; i < proc->fds->length; ++i) {
if (proc->fds->entries[i]) {
close_fs(proc->fds->entries[i]);
proc->fds->entries[i] = NULL;
}
}
debug_print(INFO, "... and their storage %d", proc->id);
free(proc->fds->entries);
free(proc->fds);
debug_print(INFO, "... and the kernel stack (hope this ain't us) %d", proc->id);
free((void *)(proc->image.stack - KERNEL_STACK_SIZE));
}
}
void reap_process(process_t * proc) {
debug_print(INFO, "Reaping process %d; mem before = %d", proc->id, memory_use());
free(proc->name);
debug_print(INFO, "Reaped process %d; mem after = %d", proc->id, memory_use());
delete_process(proc);
debug_print_process_tree();
}
static int wait_candidate(process_t * parent, int pid, int options, process_t * proc) {
if (!proc) return 0;
if (options & WNOKERN) {
/* Skip kernel processes */
if (proc->is_tasklet) return 0;
}
if (pid < -1) {
if (proc->job == -pid || proc->id == -pid) return 1;
} else if (pid == 0) {
/* Matches our group ID */
if (proc->job == parent->id) return 1;
} else if (pid > 0) {
/* Specific pid */
if (proc->id == pid) return 1;
} else {
return 1;
}
return 0;
}
int waitpid(int pid, int * status, int options) {
process_t * proc = (process_t *)current_process;
if (proc->group) {
proc = process_from_pid(proc->group);
}
debug_print(INFO, "waitpid(%s%d, ..., %d) (from pid=%d.%d)", (pid >= 0) ? "" : "-", (pid >= 0) ? pid : -pid, options, current_process->id, current_process->group);
do {
process_t * candidate = NULL;
int has_children = 0;
/* First, find out if there is anyone to reap */
foreach(node, proc->tree_entry->children) {
if (!node->value) {
continue;
}
process_t * child = ((tree_node_t *)node->value)->value;
if (wait_candidate(proc, pid, options, child)) {
has_children = 1;
if (child->finished) {
candidate = child;
break;
}
if (!(options & WEXITED) && child->suspended) {
candidate = child;
break;
}
}
}
if (!has_children) {
/* No valid children matching this description */
debug_print(INFO, "No children matching description.");
return -ECHILD;
}
if (candidate) {
debug_print(INFO, "Candidate found (%x:%d), bailing early.", candidate, candidate->id);
if (status) {
*status = candidate->status;
}
int pid = candidate->id;
if (candidate->finished) {
reap_process(candidate);
}
return pid;
} else {
if (options & WNOHANG) {
return 0;
}
debug_print(INFO, "Sleeping until queue is done.");
/* Wait */
if (sleep_on(proc->wait_queue) != 0) {
debug_print(INFO, "wait() was interrupted");
return -EINTR;
}
}
} while (1);
}
int process_wait_nodes(process_t * process,fs_node_t * nodes[], int timeout) {
assert(!process->node_waits && "Tried to wait on nodes while already waiting on nodes.");
fs_node_t ** n = nodes;
int index = 0;
if (*n) {
do {
int result = selectcheck_fs(*n);
if (result < 0) {
debug_print(NOTICE, "An invalid descriptor was specified: %d (0x%x) (pid=%d)", index, *n, current_process->id);
return -1;
}
if (result == 0) {
return index;
}
n++;
index++;
} while (*n);
}
if (timeout == 0) {
return -2;
}
n = nodes;
process->node_waits = list_create();
if (*n) {
do {
if (selectwait_fs(*n, process) < 0) {
debug_print(NOTICE, "Bad selectwait? 0x%x", *n);
}
n++;
} while (*n);
}
if (timeout > 0) {
debug_print(INFO, "fswait with a timeout of %d (pid=%d)", timeout, current_process->id);
unsigned long s, ss;
relative_time(0, timeout, &s, &ss);
IRQ_OFF;
spin_lock(sleep_lock);
node_t * before = NULL;
foreach(node, sleep_queue) {
sleeper_t * candidate = ((sleeper_t *)node->value);
if (candidate->end_tick > s || (candidate->end_tick == s && candidate->end_subtick > ss)) {
break;
}
before = node;
}
sleeper_t * proc = malloc(sizeof(sleeper_t));
proc->process = process;
proc->end_tick = s;
proc->end_subtick = ss;
proc->is_fswait = 1;
list_insert(((process_t *)process)->node_waits, proc);
process->timeout_node = list_insert_after(sleep_queue, before, proc);
spin_unlock(sleep_lock);
IRQ_RES;
} else {
process->timeout_node = NULL;
}
process->awoken_index = -1;
/* Wait. */
switch_task(0);
return process->awoken_index;
}
int process_awaken_from_fswait(process_t * process, int index) {
process->awoken_index = index;
list_free(process->node_waits);
free(process->node_waits);
process->node_waits = NULL;
if (process->timeout_node && process->timeout_node->owner == sleep_queue) {
sleeper_t * proc = process->timeout_node->value;
if (proc->is_fswait != -1) {
list_delete(sleep_queue, process->timeout_node);
free(process->timeout_node->value);
free(process->timeout_node);
}
}
process->timeout_node = NULL;
make_process_ready(process);
return 0;
}
int process_alert_node(process_t * process, void * value) {
if (!is_valid_process(process)) {
debug_print(WARNING, "Invalid process in alert from fswait.");
return 0;
}
if (!process->node_waits) {
return 0; /* Possibly already returned. Wait for another call. */
}
int index = 0;
foreach(node, process->node_waits) {
if (value == node->value) {
return process_awaken_from_fswait(process, index);
}
index++;
}
return -1;
}