haiku/src/system/kernel/port.cpp

1468 lines
32 KiB
C++
Raw Normal View History

/*
* Copyright 2002-2007, Axel Dörfler, axeld@pinc-software.de.
* Distributed under the terms of the MIT License.
*
* Copyright 2001, Mark-Jan Bastian. All rights reserved.
* Distributed under the terms of the NewOS License.
*/
/*! Ports for IPC */
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
#include <port.h>
#include <ctype.h>
#include <iovec.h>
#include <stdlib.h>
#include <string.h>
#include <OS.h>
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
#include <arch/int.h>
#include <cbuf.h>
#include <kernel.h>
#include <sem.h>
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
#include <syscall_restart.h>
#include <team.h>
#include <util/list.h>
#include <wait_for_objects.h>
//#define TRACE_PORTS
#ifdef TRACE_PORTS
# define TRACE(x) dprintf x
#else
# define TRACE(x)
#endif
typedef struct port_msg {
list_link link;
int32 code;
cbuf *buffer_chain;
size_t size;
uid_t sender;
gid_t sender_group;
team_id sender_team;
} port_msg;
struct port_entry {
port_id id;
team_id owner;
int32 capacity;
spinlock lock;
const char *name;
sem_id read_sem;
sem_id write_sem;
int32 total_count; // messages read from port since creation
select_info *select_infos;
struct list msg_queue;
};
#define MAX_QUEUE_LENGTH 4096
#define PORT_MAX_MESSAGE_SIZE 65536
// sMaxPorts must be power of 2
static int32 sMaxPorts = 4096;
static int32 sUsedPorts = 0;
static struct port_entry *sPorts = NULL;
static area_id sPortArea = 0;
static bool sPortsActive = false;
static port_id sNextPort = 1;
static int32 sFirstFreeSlot = 1;
static spinlock sPortSpinlock = 0;
#define GRAB_PORT_LIST_LOCK() acquire_spinlock(&sPortSpinlock)
#define RELEASE_PORT_LIST_LOCK() release_spinlock(&sPortSpinlock)
#define GRAB_PORT_LOCK(s) acquire_spinlock(&(s).lock)
#define RELEASE_PORT_LOCK(s) release_spinlock(&(s).lock)
static int
dump_port_list(int argc, char **argv)
{
const char *name = NULL;
team_id owner = -1;
int32 i;
if (argc > 2) {
if (!strcmp(argv[1], "team") || !strcmp(argv[1], "owner"))
owner = strtoul(argv[2], NULL, 0);
else if (!strcmp(argv[1], "name"))
name = argv[2];
} else if (argc > 1)
owner = strtoul(argv[1], NULL, 0);
kprintf("port id cap r-sem w-sem team name\n");
for (i = 0; i < sMaxPorts; i++) {
struct port_entry *port = &sPorts[i];
if (port->id < 0
|| (owner != -1 && port->owner != owner)
|| (name != NULL && strstr(port->name, name) == NULL))
continue;
kprintf("%p %6ld %4ld %6ld %6ld %6ld %s\n", port, port->id,
port->capacity, port->read_sem, port->write_sem, port->owner,
port->name);
}
return 0;
}
static void
_dump_port_info(struct port_entry *port)
{
int32 count;
kprintf("PORT: %p\n", port);
kprintf(" id: %ld\n", port->id);
kprintf(" name: \"%s\"\n", port->name);
kprintf(" owner: %ld\n", port->owner);
kprintf(" capacity: %ld\n", port->capacity);
kprintf(" read_sem: %ld\n", port->read_sem);
kprintf(" write_sem: %ld\n", port->write_sem);
get_sem_count(port->read_sem, &count);
kprintf(" read_sem count: %ld\n", count);
get_sem_count(port->write_sem, &count);
kprintf(" write_sem count: %ld\n", count);
kprintf(" total count: %ld\n", port->total_count);
set_debug_variable("_port", (addr_t)port);
set_debug_variable("_portID", port->id);
set_debug_variable("_owner", port->owner);
set_debug_variable("_readSem", port->read_sem);
set_debug_variable("_writeSem", port->write_sem);
}
static int
dump_port_info(int argc, char **argv)
{
const char *name = NULL;
sem_id sem = -1;
int i;
if (argc < 2) {
print_debugger_command_usage(argv[0]);
return 0;
}
if (argc > 2) {
if (!strcmp(argv[1], "address")) {
_dump_port_info((struct port_entry *)strtoul(argv[2], NULL, 0));
return 0;
} else if (!strcmp(argv[1], "sem"))
sem = strtoul(argv[2], NULL, 0);
else if (!strcmp(argv[1], "name"))
name = argv[2];
} else if (isdigit(argv[1][0])) {
// if the argument looks like a number, treat it as such
uint32 num = strtoul(argv[1], NULL, 0);
uint32 slot = num % sMaxPorts;
if (sPorts[slot].id != (int)num) {
kprintf("port %ld (%#lx) doesn't exist!\n", num, num);
return 0;
}
_dump_port_info(&sPorts[slot]);
return 0;
} else
name = argv[1];
// walk through the ports list, trying to match name
for (i = 0; i < sMaxPorts; i++) {
if ((name != NULL && sPorts[i].name != NULL
&& !strcmp(name, sPorts[i].name))
|| (sem != -1 && (sPorts[i].read_sem == sem
|| sPorts[i].write_sem == sem))) {
_dump_port_info(&sPorts[i]);
return 0;
}
}
return 0;
}
static void
notify_port_select_events(int slot, uint16 events)
{
if (sPorts[slot].select_infos)
notify_select_events_list(sPorts[slot].select_infos, events);
}
static void
put_port_msg(port_msg *msg)
{
cbuf_free_chain(msg->buffer_chain);
free(msg);
}
static port_msg *
get_port_msg(int32 code, size_t bufferSize)
{
// ToDo: investigate preallocation of port_msgs (or use a slab allocator)
cbuf *bufferChain = NULL;
port_msg *msg = (port_msg *)malloc(sizeof(port_msg));
if (msg == NULL)
return NULL;
if (bufferSize > 0) {
bufferChain = cbuf_get_chain(bufferSize);
if (bufferChain == NULL) {
free(msg);
return NULL;
}
}
msg->code = code;
msg->buffer_chain = bufferChain;
msg->size = bufferSize;
return msg;
}
/*! You need to own the port's lock when calling this function */
static bool
is_port_closed(int32 slot)
{
return sPorts[slot].capacity == 0;
}
/*! Fills the port_info structure with information from the specified
port.
The port lock must be held when called.
*/
static void
fill_port_info(struct port_entry *port, port_info *info, size_t size)
{
int32 count;
info->port = port->id;
info->team = port->owner;
info->capacity = port->capacity;
get_sem_count(port->read_sem, &count);
if (count < 0)
count = 0;
info->queue_count = count;
info->total_count = port->total_count;
strlcpy(info->name, port->name, B_OS_NAME_LENGTH);
}
// #pragma mark - private kernel API
/*! This function cycles through the ports table, deleting all
the ports that are owned by the passed team_id
*/
int
delete_owned_ports(team_id owner)
{
// ToDo: investigate maintaining a list of ports in the team
// to make this simpler and more efficient.
cpu_status state;
int i;
int count = 0;
TRACE(("delete_owned_ports(owner = %ld)\n", owner));
if (!sPortsActive)
return B_BAD_PORT_ID;
state = disable_interrupts();
GRAB_PORT_LIST_LOCK();
for (i = 0; i < sMaxPorts; i++) {
if (sPorts[i].id != -1 && sPorts[i].owner == owner) {
port_id id = sPorts[i].id;
RELEASE_PORT_LIST_LOCK();
restore_interrupts(state);
delete_port(id);
count++;
state = disable_interrupts();
GRAB_PORT_LIST_LOCK();
}
}
RELEASE_PORT_LIST_LOCK();
restore_interrupts(state);
return count;
}
int32
port_max_ports(void)
{
return sMaxPorts;
}
int32
port_used_ports(void)
{
return sUsedPorts;
}
status_t
port_init(kernel_args *args)
{
size_t size = sizeof(struct port_entry) * sMaxPorts;
int32 i;
// create and initialize ports table
sPortArea = create_area("port_table", (void **)&sPorts, B_ANY_KERNEL_ADDRESS,
size, B_FULL_LOCK, B_KERNEL_READ_AREA | B_KERNEL_WRITE_AREA);
if (sPortArea < 0) {
panic("unable to allocate kernel port table!\n");
return sPortArea;
}
// ToDo: investigate preallocating a list of port_msgs to
// speed up actual message sending/receiving, a slab allocator
// might do it as well, though :-)
memset(sPorts, 0, size);
for (i = 0; i < sMaxPorts; i++)
sPorts[i].id = -1;
// add debugger commands
add_debugger_command_etc("ports", &dump_port_list,
"Dump a list of all active ports (for team, with name, etc.)",
"[ ([ \"team\" | \"owner\" ] <team>) | (\"name\" <name>) ]\n"
"Prints a list of all active ports meeting the given\n"
"requirement. If no argument is given, all ports are listed.\n"
" <team> - The team owning the ports.\n"
" <name> - Part of the name of the ports.\n", 0);
add_debugger_command_etc("port", &dump_port_info,
"Dump info about a particular port",
"([ \"address\" ] <address>) | ([ \"name\" ] <name>) "
"| (\"sem\" <sem>)\n"
"Prints info about the specified port.\n"
" <address> - Pointer to the port structure.\n"
" <name> - Name of the port.\n"
" <sem> - ID of the port's read or write semaphore.\n", 0);
sPortsActive = true;
return B_OK;
}
// #pragma mark - public kernel API
port_id
create_port(int32 queueLength, const char *name)
{
cpu_status state;
char nameBuffer[B_OS_NAME_LENGTH];
sem_id readSem, writeSem;
status_t status;
team_id owner;
int32 slot;
TRACE(("create_port(queueLength = %ld, name = \"%s\")\n", queueLength, name));
if (!sPortsActive)
return B_BAD_PORT_ID;
// check queue length
if (queueLength < 1
|| queueLength > MAX_QUEUE_LENGTH)
return B_BAD_VALUE;
// check early on if there are any free port slots to use
if (atomic_add(&sUsedPorts, 1) >= sMaxPorts) {
status = B_NO_MORE_PORTS;
goto err1;
}
// check & dup name
if (name == NULL)
name = "unnamed port";
// ToDo: we could save the memory and use the semaphore name only instead
strlcpy(nameBuffer, name, B_OS_NAME_LENGTH);
name = strdup(nameBuffer);
if (name == NULL) {
status = B_NO_MEMORY;
goto err1;
}
// create read sem with owner set to -1
// ToDo: should be B_SYSTEM_TEAM
readSem = create_sem_etc(0, name, -1);
if (readSem < B_OK) {
status = readSem;
goto err2;
}
// create write sem
writeSem = create_sem_etc(queueLength, name, -1);
if (writeSem < B_OK) {
status = writeSem;
goto err3;
}
owner = team_get_current_team_id();
state = disable_interrupts();
GRAB_PORT_LIST_LOCK();
// find the first empty spot
for (slot = 0; slot < sMaxPorts; slot++) {
int32 i = (slot + sFirstFreeSlot) % sMaxPorts;
if (sPorts[i].id == -1) {
port_id id;
// make the port_id be a multiple of the slot it's in
if (i >= sNextPort % sMaxPorts)
sNextPort += i - sNextPort % sMaxPorts;
else
sNextPort += sMaxPorts - (sNextPort % sMaxPorts - i);
sFirstFreeSlot = slot + 1;
GRAB_PORT_LOCK(sPorts[i]);
sPorts[i].id = sNextPort++;
RELEASE_PORT_LIST_LOCK();
sPorts[i].capacity = queueLength;
sPorts[i].owner = owner;
sPorts[i].name = name;
sPorts[i].read_sem = readSem;
sPorts[i].write_sem = writeSem;
list_init(&sPorts[i].msg_queue);
sPorts[i].total_count = 0;
sPorts[i].select_infos = NULL;
id = sPorts[i].id;
RELEASE_PORT_LOCK(sPorts[i]);
restore_interrupts(state);
TRACE(("create_port() done: port created %ld\n", id));
return id;
}
}
// not enough ports...
// ToDo: due to sUsedPorts, this cannot happen anymore - as
// long as sMaxPorts stays constant over the kernel run
// time (which it should be). IOW we could simply panic()
// here.
RELEASE_PORT_LIST_LOCK();
restore_interrupts(state);
status = B_NO_MORE_PORTS;
delete_sem(writeSem);
err3:
delete_sem(readSem);
err2:
free((char *)name);
err1:
atomic_add(&sUsedPorts, -1);
return status;
}
status_t
close_port(port_id id)
{
sem_id readSem, writeSem;
cpu_status state;
int32 slot;
TRACE(("close_port(id = %ld)\n", id));
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
slot = id % sMaxPorts;
// walk through the sem list, trying to match name
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("close_port: invalid port_id %ld\n", id));
return B_BAD_PORT_ID;
}
// mark port to disable writing - deleting the semaphores will
// wake up waiting read/writes
sPorts[slot].capacity = 0;
readSem = sPorts[slot].read_sem;
writeSem = sPorts[slot].write_sem;
notify_port_select_events(slot, B_EVENT_INVALID);
sPorts[slot].select_infos = NULL;
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
delete_sem(readSem);
delete_sem(writeSem);
return B_NO_ERROR;
}
status_t
delete_port(port_id id)
{
cpu_status state;
sem_id readSem, writeSem;
const char *name;
struct list list;
port_msg *msg;
int32 slot;
TRACE(("delete_port(id = %ld)\n", id));
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("delete_port: invalid port_id %ld\n", id));
return B_BAD_PORT_ID;
}
/* mark port as invalid */
sPorts[slot].id = -1;
name = sPorts[slot].name;
readSem = sPorts[slot].read_sem;
writeSem = sPorts[slot].write_sem;
sPorts[slot].name = NULL;
list_move_to_list(&sPorts[slot].msg_queue, &list);
notify_port_select_events(slot, B_EVENT_INVALID);
sPorts[slot].select_infos = NULL;
RELEASE_PORT_LOCK(sPorts[slot]);
// update the first free slot hint in the array
GRAB_PORT_LIST_LOCK();
if (slot < sFirstFreeSlot)
sFirstFreeSlot = slot;
RELEASE_PORT_LIST_LOCK();
restore_interrupts(state);
atomic_add(&sUsedPorts, -1);
// free the queue
while ((msg = (port_msg *)list_remove_head_item(&list)) != NULL) {
put_port_msg(msg);
}
free((char *)name);
// release the threads that were blocking on this port by deleting the sem
// read_port() will see the B_BAD_SEM_ID acq_sem() return value, and act accordingly
delete_sem(readSem);
delete_sem(writeSem);
return B_OK;
}
status_t
select_port(int32 id, struct select_info *info, bool kernel)
{
cpu_status state;
int32 slot;
status_t error = B_OK;
if (id < 0)
return B_BAD_PORT_ID;
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id || is_port_closed(slot)) {
// bad port ID
error = B_BAD_SEM_ID;
} else if (!kernel && sPorts[slot].owner == team_get_kernel_team_id()) {
// kernel port, but call from userland
error = B_NOT_ALLOWED;
} else {
info->selected_events &= B_EVENT_READ | B_EVENT_WRITE | B_EVENT_INVALID;
if (info->selected_events != 0) {
uint16 events = 0;
int32 writeCount = 0;
info->next = sPorts[slot].select_infos;
sPorts[slot].select_infos = info;
// check for events
if ((info->selected_events & B_EVENT_READ) != 0
&& !list_is_empty(&sPorts[slot].msg_queue)) {
events |= B_EVENT_READ;
}
if (get_sem_count(sPorts[slot].write_sem, &writeCount) == B_OK
&& writeCount > 0) {
events |= B_EVENT_WRITE;
}
if (events != 0)
notify_select_events(info, events);
}
}
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
return error;
}
status_t
deselect_port(int32 id, struct select_info *info, bool kernel)
{
cpu_status state;
int32 slot;
if (id < 0)
return B_BAD_PORT_ID;
if (info->selected_events == 0)
return B_OK;
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id == id) {
select_info** infoLocation = &sPorts[slot].select_infos;
while (*infoLocation != NULL && *infoLocation != info)
infoLocation = &(*infoLocation)->next;
if (*infoLocation == info)
*infoLocation = info->next;
}
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
return B_OK;
}
port_id
find_port(const char *name)
{
port_id portFound = B_NAME_NOT_FOUND;
cpu_status state;
int32 i;
TRACE(("find_port(name = \"%s\")\n", name));
if (!sPortsActive)
return B_NAME_NOT_FOUND;
if (name == NULL)
return B_BAD_VALUE;
// Since we have to check every single port, and we don't
// care if it goes away at any point, we're only grabbing
// the port lock in question, not the port list lock
// loop over list
for (i = 0; i < sMaxPorts && portFound < B_OK; i++) {
// lock every individual port before comparing
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[i]);
if (sPorts[i].id >= 0 && !strcmp(name, sPorts[i].name))
portFound = sPorts[i].id;
RELEASE_PORT_LOCK(sPorts[i]);
restore_interrupts(state);
}
return portFound;
}
status_t
_get_port_info(port_id id, port_info *info, size_t size)
{
cpu_status state;
int slot;
TRACE(("get_port_info(id = %ld)\n", id));
if (info == NULL || size != sizeof(port_info))
return B_BAD_VALUE;
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id || sPorts[slot].capacity == 0) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("get_port_info: invalid port_id %ld\n", id));
return B_BAD_PORT_ID;
}
// fill a port_info struct with info
fill_port_info(&sPorts[slot], info, size);
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
return B_OK;
}
status_t
_get_next_port_info(team_id team, int32 *_cookie, struct port_info *info, size_t size)
{
cpu_status state;
int slot;
TRACE(("get_next_port_info(team = %ld)\n", team));
if (info == NULL || size != sizeof(port_info) || _cookie == NULL || team < B_OK)
return B_BAD_VALUE;
if (!sPortsActive)
return B_BAD_PORT_ID;
slot = *_cookie;
if (slot >= sMaxPorts)
return B_BAD_PORT_ID;
if (team == B_CURRENT_TEAM)
team = team_get_current_team_id();
info->port = -1; // used as found flag
// spinlock
state = disable_interrupts();
GRAB_PORT_LIST_LOCK();
while (slot < sMaxPorts) {
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != -1 && sPorts[slot].capacity != 0 && sPorts[slot].owner == team) {
// found one!
fill_port_info(&sPorts[slot], info, size);
RELEASE_PORT_LOCK(sPorts[slot]);
slot++;
break;
}
RELEASE_PORT_LOCK(sPorts[slot]);
slot++;
}
RELEASE_PORT_LIST_LOCK();
restore_interrupts(state);
if (info->port == -1)
return B_BAD_PORT_ID;
*_cookie = slot;
return B_NO_ERROR;
}
ssize_t
port_buffer_size(port_id id)
{
return port_buffer_size_etc(id, 0, 0);
}
ssize_t
port_buffer_size_etc(port_id id, uint32 flags, bigtime_t timeout)
{
port_message_info info;
status_t error = get_port_message_info_etc(id, &info, flags, timeout);
return error != B_OK ? error : info.size;
}
status_t
_get_port_message_info_etc(port_id id, port_message_info *info,
size_t infoSize, uint32 flags, bigtime_t timeout)
{
if (info == NULL || infoSize != sizeof(port_message_info))
return B_BAD_VALUE;
cpu_status state;
sem_id cachedSem;
status_t status;
port_msg *msg;
int32 slot;
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id
|| (is_port_closed(slot) && list_is_empty(&sPorts[slot].msg_queue))) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("port_buffer_size_etc(): %s port %ld\n",
sPorts[slot].id == id ? "closed" : "invalid", id));
return B_BAD_PORT_ID;
}
cachedSem = sPorts[slot].read_sem;
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
// block if no message, or, if B_TIMEOUT flag set, block with timeout
status = acquire_sem_etc(cachedSem, 1, flags, timeout);
if (status != B_OK && status != B_BAD_SEM_ID)
return status;
// in case of B_BAD_SEM_ID, the port might have been closed but not yet
// deleted, ie. there could still be messages waiting for us
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id) {
// the port is no longer there
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
return B_BAD_PORT_ID;
}
// determine tail & get the length of the message
status_t error = B_OK;
msg = (port_msg*)list_get_first_item(&sPorts[slot].msg_queue);
if (msg == NULL) {
if (status == B_OK)
panic("port %ld: no messages found\n", sPorts[slot].id);
error = B_BAD_PORT_ID;
} else {
info->size = msg->size;
info->sender = msg->sender;
info->sender_group = msg->sender_group;
info->sender_team = msg->sender_team;
}
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
// restore read_sem, as we haven't read from the port
release_sem(cachedSem);
// return length of item at end of queue
return error;
}
ssize_t
port_count(port_id id)
{
cpu_status state;
int32 count = 0;
int32 slot;
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("port_count: invalid port_id %ld\n", id));
return B_BAD_PORT_ID;
}
if (get_sem_count(sPorts[slot].read_sem, &count) == B_OK) {
// do not return negative numbers
if (count < 0)
count = 0;
} else {
// the port might have been closed - we need to actually count the messages
void *message = NULL;
while ((message = list_get_next_item(&sPorts[slot].msg_queue, message)) != NULL) {
count++;
}
}
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
// return count of messages
return count;
}
ssize_t
read_port(port_id port, int32 *msgCode, void *msgBuffer, size_t bufferSize)
{
return read_port_etc(port, msgCode, msgBuffer, bufferSize, 0, 0);
}
ssize_t
read_port_etc(port_id id, int32 *_msgCode, void *msgBuffer, size_t bufferSize,
uint32 flags, bigtime_t timeout)
{
cpu_status state;
sem_id cachedSem;
status_t status;
bool userCopy = (flags & PORT_FLAG_USE_USER_MEMCPY) > 0;
port_msg *msg;
size_t size;
int slot;
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
if ((msgBuffer == NULL && bufferSize > 0)
|| timeout < 0)
return B_BAD_VALUE;
flags = flags & (B_CAN_INTERRUPT | B_KILL_CAN_INTERRUPT
| B_RELATIVE_TIMEOUT | B_ABSOLUTE_TIMEOUT);
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id
|| (is_port_closed(slot) && list_is_empty(&sPorts[slot].msg_queue))) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("read_port_etc(): %s port %ld\n",
sPorts[slot].id == id ? "closed" : "invalid", id));
return B_BAD_PORT_ID;
}
// store sem_id in local variable
cachedSem = sPorts[slot].read_sem;
// unlock port && enable ints/
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
status = acquire_sem_etc(cachedSem, 1, flags, timeout);
// get 1 entry from the queue, block if needed
if (status != B_OK && status != B_BAD_SEM_ID)
return status;
// in case of B_BAD_SEM_ID, the port might have been closed but not yet
// deleted, ie. there could still be messages waiting for us
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
// first, let's check if the port is still alive
if (sPorts[slot].id == -1) {
// the port has been deleted in the meantime
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
return B_BAD_PORT_ID;
}
msg = (port_msg*)list_get_first_item(&sPorts[slot].msg_queue);
if (msg == NULL) {
if (status == B_OK)
panic("port %ld: no messages found", sPorts[slot].id);
// the port has obviously been closed, but no messages are left anymore
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
return B_BAD_PORT_ID;
}
list_remove_link(msg);
sPorts[slot].total_count++;
notify_port_select_events(slot, B_EVENT_WRITE);
cachedSem = sPorts[slot].write_sem;
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
// check output buffer size
size = min_c(bufferSize, msg->size);
// copy message
if (_msgCode != NULL)
*_msgCode = msg->code;
if (size > 0) {
if (userCopy) {
if ((status = cbuf_user_memcpy_from_chain(msgBuffer, msg->buffer_chain, 0, size) < B_OK)) {
// leave the port intact, for other threads that might not crash
put_port_msg(msg);
release_sem(cachedSem);
return status;
}
} else
cbuf_memcpy_from_chain(msgBuffer, msg->buffer_chain, 0, size);
}
put_port_msg(msg);
// make one spot in queue available again for write
release_sem(cachedSem);
// ToDo: we might think about setting B_NO_RESCHEDULE here
// from time to time (always?)
return size;
}
status_t
write_port(port_id id, int32 msgCode, const void *msgBuffer, size_t bufferSize)
{
iovec vec = { (void *)msgBuffer, bufferSize };
return writev_port_etc(id, msgCode, &vec, 1, bufferSize, 0, 0);
}
status_t
write_port_etc(port_id id, int32 msgCode, const void *msgBuffer,
size_t bufferSize, uint32 flags, bigtime_t timeout)
{
iovec vec = { (void *)msgBuffer, bufferSize };
return writev_port_etc(id, msgCode, &vec, 1, bufferSize, flags, timeout);
}
status_t
writev_port_etc(port_id id, int32 msgCode, const iovec *msgVecs,
size_t vecCount, size_t bufferSize, uint32 flags,
bigtime_t timeout)
{
cpu_status state;
sem_id cachedSem;
status_t status;
port_msg *msg;
bool userCopy = (flags & PORT_FLAG_USE_USER_MEMCPY) > 0;
int slot;
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
// mask irrelevant flags (for acquire_sem() usage)
flags = flags & (B_CAN_INTERRUPT | B_KILL_CAN_INTERRUPT
| B_RELATIVE_TIMEOUT | B_ABSOLUTE_TIMEOUT);
slot = id % sMaxPorts;
if (bufferSize > PORT_MAX_MESSAGE_SIZE)
return B_BAD_VALUE;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("write_port_etc: invalid port_id %ld\n", id));
return B_BAD_PORT_ID;
}
if (is_port_closed(slot)) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("write_port_etc: port %ld closed\n", id));
return B_BAD_PORT_ID;
}
// store sem_id in local variable
cachedSem = sPorts[slot].write_sem;
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
status = acquire_sem_etc(cachedSem, 1, flags, timeout);
// get 1 entry from the queue, block if needed
if (status == B_BAD_SEM_ID) {
// somebody deleted or closed the port
return B_BAD_PORT_ID;
}
if (status != B_OK)
return status;
msg = get_port_msg(msgCode, bufferSize);
if (msg == NULL)
return B_NO_MEMORY;
// sender credentials
msg->sender = geteuid();
msg->sender_group = getegid();
msg->sender_team = team_get_current_team_id();
if (bufferSize > 0) {
uint32 i;
if (userCopy) {
// copy from user memory
for (i = 0; i < vecCount; i++) {
size_t bytes = msgVecs[i].iov_len;
if (bytes > bufferSize)
bytes = bufferSize;
if ((status = cbuf_user_memcpy_to_chain(msg->buffer_chain,
0, msgVecs[i].iov_base, bytes)) < B_OK)
return status;
bufferSize -= bytes;
if (bufferSize == 0)
break;
}
} else {
// copy from kernel memory
for (i = 0; i < vecCount; i++) {
size_t bytes = msgVecs[i].iov_len;
if (bytes > bufferSize)
bytes = bufferSize;
if ((status = cbuf_memcpy_to_chain(msg->buffer_chain,
0, msgVecs[i].iov_base, bytes)) < 0)
return status;
bufferSize -= bytes;
if (bufferSize == 0)
break;
}
}
}
// attach message to queue
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
// first, let's check if the port is still alive
if (sPorts[slot].id == -1) {
// the port has been deleted in the meantime
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
put_port_msg(msg);
return B_BAD_PORT_ID;
}
list_add_item(&sPorts[slot].msg_queue, msg);
notify_port_select_events(slot, B_EVENT_READ);
// store sem_id in local variable
cachedSem = sPorts[slot].read_sem;
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
// release sem, allowing read (might reschedule)
release_sem(cachedSem);
return B_NO_ERROR;
}
status_t
set_port_owner(port_id id, team_id team)
{
cpu_status state;
int slot;
// ToDo: Shouldn't we at least check, whether the team exists?
TRACE(("set_port_owner(id = %ld, team = %ld)\n", id, team));
if (!sPortsActive || id < 0)
return B_BAD_PORT_ID;
slot = id % sMaxPorts;
state = disable_interrupts();
GRAB_PORT_LOCK(sPorts[slot]);
if (sPorts[slot].id != id) {
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
TRACE(("set_port_owner: invalid port_id %ld\n", id));
return B_BAD_PORT_ID;
}
// transfer ownership to other team
sPorts[slot].owner = team;
// unlock port
RELEASE_PORT_LOCK(sPorts[slot]);
restore_interrupts(state);
return B_NO_ERROR;
}
// #pragma mark - syscalls
port_id
_user_create_port(int32 queueLength, const char *userName)
{
char name[B_OS_NAME_LENGTH];
if (userName == NULL)
return create_port(queueLength, NULL);
if (!IS_USER_ADDRESS(userName)
|| user_strlcpy(name, userName, B_OS_NAME_LENGTH) < B_OK)
return B_BAD_ADDRESS;
return create_port(queueLength, name);
}
status_t
_user_close_port(port_id id)
{
return close_port(id);
}
status_t
_user_delete_port(port_id id)
{
return delete_port(id);
}
port_id
_user_find_port(const char *userName)
{
char name[B_OS_NAME_LENGTH];
if (userName == NULL)
return B_BAD_VALUE;
if (!IS_USER_ADDRESS(userName)
|| user_strlcpy(name, userName, B_OS_NAME_LENGTH) < B_OK)
return B_BAD_ADDRESS;
return find_port(name);
}
status_t
_user_get_port_info(port_id id, struct port_info *userInfo)
{
struct port_info info;
status_t status;
if (userInfo == NULL)
return B_BAD_VALUE;
if (!IS_USER_ADDRESS(userInfo))
return B_BAD_ADDRESS;
status = get_port_info(id, &info);
// copy back to user space
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
if (status == B_OK
&& user_memcpy(userInfo, &info, sizeof(struct port_info)) < B_OK)
return B_BAD_ADDRESS;
return status;
}
status_t
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
_user_get_next_port_info(team_id team, int32 *userCookie,
struct port_info *userInfo)
{
struct port_info info;
status_t status;
int32 cookie;
if (userCookie == NULL || userInfo == NULL)
return B_BAD_VALUE;
if (!IS_USER_ADDRESS(userCookie) || !IS_USER_ADDRESS(userInfo)
|| user_memcpy(&cookie, userCookie, sizeof(int32)) < B_OK)
return B_BAD_ADDRESS;
status = get_next_port_info(team, &cookie, &info);
// copy back to user space
if (user_memcpy(userCookie, &cookie, sizeof(int32)) < B_OK
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
|| (status == B_OK && user_memcpy(userInfo, &info,
sizeof(struct port_info)) < B_OK))
return B_BAD_ADDRESS;
return status;
}
ssize_t
_user_port_buffer_size_etc(port_id port, uint32 flags, bigtime_t timeout)
{
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
syscall_restart_handle_timeout_pre(flags, timeout);
status_t status = port_buffer_size_etc(port, flags | B_CAN_INTERRUPT,
timeout);
return syscall_restart_handle_timeout_post(status, timeout);
}
ssize_t
_user_port_count(port_id port)
{
return port_count(port);
}
status_t
_user_set_port_owner(port_id port, team_id team)
{
return set_port_owner(port, team);
}
ssize_t
_user_read_port_etc(port_id port, int32 *userCode, void *userBuffer,
size_t bufferSize, uint32 flags, bigtime_t timeout)
{
int32 messageCode;
ssize_t bytesRead;
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
syscall_restart_handle_timeout_pre(flags, timeout);
if (userBuffer == NULL && bufferSize != 0)
return B_BAD_VALUE;
if ((userCode != NULL && !IS_USER_ADDRESS(userCode))
|| (userBuffer != NULL && !IS_USER_ADDRESS(userBuffer)))
return B_BAD_ADDRESS;
bytesRead = read_port_etc(port, &messageCode, userBuffer, bufferSize,
flags | PORT_FLAG_USE_USER_MEMCPY | B_CAN_INTERRUPT, timeout);
if (bytesRead >= 0 && userCode != NULL
&& user_memcpy(userCode, &messageCode, sizeof(int32)) < B_OK)
return B_BAD_ADDRESS;
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
return syscall_restart_handle_timeout_post(bytesRead, timeout);
}
status_t
_user_write_port_etc(port_id port, int32 messageCode, const void *userBuffer,
size_t bufferSize, uint32 flags, bigtime_t timeout)
{
iovec vec = { (void *)userBuffer, bufferSize };
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
syscall_restart_handle_timeout_pre(flags, timeout);
if (userBuffer == NULL && bufferSize != 0)
return B_BAD_VALUE;
if (userBuffer != NULL && !IS_USER_ADDRESS(userBuffer))
return B_BAD_ADDRESS;
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
status_t status = writev_port_etc(port, messageCode, &vec, 1, bufferSize,
flags | PORT_FLAG_USE_USER_MEMCPY | B_CAN_INTERRUPT, timeout);
return syscall_restart_handle_timeout_post(status, timeout);
}
status_t
_user_writev_port_etc(port_id port, int32 messageCode, const iovec *userVecs,
size_t vecCount, size_t bufferSize, uint32 flags, bigtime_t timeout)
{
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
syscall_restart_handle_timeout_pre(flags, timeout);
if (userVecs == NULL && bufferSize != 0)
return B_BAD_VALUE;
if (userVecs != NULL && !IS_USER_ADDRESS(userVecs))
return B_BAD_ADDRESS;
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
iovec *vecs = NULL;
if (userVecs && vecCount != 0) {
vecs = (iovec*)malloc(sizeof(iovec) * vecCount);
if (vecs == NULL)
return B_NO_MEMORY;
if (user_memcpy(vecs, userVecs, sizeof(iovec) * vecCount) < B_OK) {
free(vecs);
return B_BAD_ADDRESS;
}
}
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
status_t status = writev_port_etc(port, messageCode, vecs, vecCount,
bufferSize, flags | PORT_FLAG_USE_USER_MEMCPY | B_CAN_INTERRUPT,
timeout);
free(vecs);
axeld + bonefish: * Implemented automatic syscall restarts: - A syscall can indicate that it has been interrupted and can be restarted by setting a respective bit in thread::flags. It can store parameters it wants to be preserved for the restart in thread::syscall_restart::parameters. Another thread::flags bit indicates whether it has been restarted. - handle_signals() clears the restart flag, if the handled signal has a handler function installed and SA_RESTART is not set. Another thread flag (THREAD_FLAGS_DONT_RESTART_SYSCALL) can prevent syscalls from being restarted, even if they could be (not used yet, but we might want to use it in resume_thread(), so that we stay behaviorally compatible with BeOS). - The architecture specific syscall handler restarts the syscall, if the restart flag is set. Implemented for x86 only. - Added some support functions in the private <syscall_restart.h> to simplify the syscall restart code in the syscalls. - Adjusted all syscalls that can potentially be restarted accordingly. - _user_ioctl() sets new thread flag THREAD_FLAGS_IOCTL_SYSCALL while calling the underlying FS's/driver's hook, so that syscall restarts can also be supported there. * thread_at_kernel_exit() invokes handle_signals() in a loop now, as long as the latter indicates that the thread shall be suspended, so that after waking up signals received in the meantime will be handled before the thread returns to userland. Adjusted handle_signals() accordingly -- when encountering a suspending signal we don't check for further signals. * Fixed sigsuspend(): Suspending the thread and rescheduling doesn't result in the correct behavior. Instead we employ a temporary condition variable and interruptably wait on it. The POSIX test suite test passes, now. * Made the switch_sem[_etc]() behavior on interruption consistent. Depending on when the signal arrived (before the call or when already waiting) the first semaphore would or wouldn't be released. Now we consistently release it. * Refactored _user_{read,write}[v]() syscalls. Use a common function for either pair. The iovec version doesn't fail anymore, if anything could be read/written at all. It also checks whether a complete vector could be read/written, so that we won't skip data, if the underlying FS/driver couldn't read/write more ATM. * Some refactoring in the x86 syscall handler: The int 99 and sysenter handlers use a common subroutine to avoid code duplication. git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23983 a95241bf-73f2-0310-859d-f6bbb57e9c96
2008-02-17 18:48:30 +03:00
return syscall_restart_handle_timeout_post(status, timeout);
}
status_t
_user_get_port_message_info_etc(port_id port, port_message_info *userInfo,
size_t infoSize, uint32 flags, bigtime_t timeout)
{
if (userInfo == NULL || infoSize != sizeof(port_message_info))
return B_BAD_VALUE;
syscall_restart_handle_timeout_pre(flags, timeout);
port_message_info info;
status_t error = _get_port_message_info_etc(port, &info, sizeof(info),
flags | B_CAN_INTERRUPT, timeout);
// copy info to userland
if (error == B_OK && (!IS_USER_ADDRESS(userInfo)
|| user_memcpy(userInfo, &info, sizeof(info)) != B_OK)) {
error = B_BAD_ADDRESS;
}
return syscall_restart_handle_timeout_post(error, timeout);
}