cosmopolitan/libc/calls/read-nt.c

749 lines
22 KiB
C
Raw Normal View History

2020-06-15 14:18:57 +00:00
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
vi: set net ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi
Copyright 2020 Justine Alexandra Roberts Tunney
2020-12-28 01:18:44 +00:00
Permission to use, copy, modify, and/or distribute this software for
any purpose with or without fee is hereby granted, provided that the
above copyright notice and this permission notice appear in all copies.
2020-06-15 14:18:57 +00:00
2020-12-28 01:18:44 +00:00
THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL
WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE
AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
PERFORMANCE OF THIS SOFTWARE.
2020-06-15 14:18:57 +00:00
*/
#include "libc/calls/internal.h"
#include "libc/calls/sig.internal.h"
#include "libc/calls/struct/fd.internal.h"
#include "libc/calls/struct/iovec.h"
2022-05-23 22:06:11 +00:00
#include "libc/calls/syscall_support-nt.internal.h"
#include "libc/calls/ttydefaults.h"
#include "libc/errno.h"
#include "libc/fmt/itoa.h"
#include "libc/intrin/atomic.h"
#include "libc/intrin/dll.h"
#include "libc/intrin/kprintf.h"
#include "libc/intrin/nomultics.internal.h"
#include "libc/intrin/strace.internal.h"
#include "libc/intrin/weaken.h"
2023-07-23 17:57:18 +00:00
#include "libc/macros.internal.h"
#include "libc/mem/mem.h"
#include "libc/nt/console.h"
#include "libc/nt/enum/consolemodeflags.h"
#include "libc/nt/enum/filetype.h"
#include "libc/nt/enum/vk.h"
#include "libc/nt/enum/wait.h"
2020-06-15 14:18:57 +00:00
#include "libc/nt/errors.h"
#include "libc/nt/events.h"
#include "libc/nt/files.h"
2020-06-15 14:18:57 +00:00
#include "libc/nt/runtime.h"
#include "libc/nt/struct/inputrecord.h"
#include "libc/nt/synchronization.h"
#include "libc/nt/thread.h"
#include "libc/nt/thunk/msabi.h"
#include "libc/str/str.h"
#include "libc/str/utf16.h"
#include "libc/sysv/consts/o.h"
#include "libc/sysv/consts/sig.h"
#include "libc/sysv/consts/termios.h"
2020-06-15 14:18:57 +00:00
#include "libc/sysv/errfuns.h"
#include "libc/thread/thread.h"
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
#include "libc/thread/tls.h"
#ifdef __x86_64__
static const struct {
int vk;
int normal_str;
int shift_str;
int ctrl_str;
int shift_ctrl_str;
} kVirtualKey[] = {
#define S(s) W(s "\0\0")
#define W(s) (s[3] << 24 | s[2] << 16 | s[1] << 8 | s[0])
{kNtVkUp, S("A"), S("1;2A"), S("1;5A"), S("1;6A")},
{kNtVkDown, S("B"), S("1;2B"), S("1;5B"), S("1;6B")},
{kNtVkLeft, S("D"), S("1;2D"), S("1;5D"), S("1;6D")},
{kNtVkRight, S("C"), S("1;2C"), S("1;5C"), S("1;6C")},
{kNtVkInsert, S("2~"), S("2;2~"), S("2;5~"), S("2;6~")},
{kNtVkDelete, S("3~"), S("3;2~"), S("3;5~"), S("3;6~")},
{kNtVkHome, S("H"), S("1;2H"), S("1;5H"), S("1;6H")},
{kNtVkEnd, S("F"), S("1;2F"), S("1;5F"), S("1;6F")},
{kNtVkPrior, S("5~"), S("5;2~"), S("5;5~"), S("5;6~")},
{kNtVkNext, S("6~"), S("6;2~"), S("6;5~"), S("6;6~")},
{kNtVkF1, -S("OP"), S("1;2P"), S("11^"), S("1;6P")},
{kNtVkF2, -S("OQ"), S("1;2Q"), S("12^"), S("1;6Q")},
{kNtVkF3, -S("OR"), S("1;2R"), S("13^"), S("1;6R")},
{kNtVkF4, -S("OS"), S("1;2S"), S("14^"), S("1;6S")},
{kNtVkF5, S("15~"), S("28~"), S("15^"), S("28^")},
{kNtVkF6, S("17~"), S("29~"), S("17^"), S("29^")},
{kNtVkF7, S("18~"), S("31~"), S("18^"), S("31^")},
{kNtVkF8, S("19~"), S("32~"), S("19^"), S("32^")},
{kNtVkF9, S("20~"), S("33~"), S("20^"), S("33^")},
{kNtVkF10, S("21~"), S("34~"), S("21^"), S("34^")},
{kNtVkF11, S("23~"), S("23$"), S("23^"), S("23@")},
{kNtVkF12, S("24~"), S("24$"), S("24^"), S("24@")},
#undef W
#undef S
};
#define KEYSTROKE_CONTAINER(e) DLL_CONTAINER(struct Keystroke, elem, e)
struct Keystroke {
char buf[23];
unsigned char buflen;
struct Dll elem;
};
struct Keystrokes {
struct Dll *list;
struct Dll *line;
struct Dll *free;
bool end_of_file;
unsigned char pc;
uint16_t utf16hs;
pthread_mutex_t lock;
2023-10-04 06:53:59 +00:00
struct Keystroke pool[512];
};
static struct Keystrokes __keystroke;
static textwindows void LockKeystrokes(void) {
pthread_mutex_lock(&__keystroke.lock);
}
static textwindows void UnlockKeystrokes(void) {
pthread_mutex_unlock(&__keystroke.lock);
}
static textwindows uint64_t BlockSignals(void) {
return atomic_exchange(&__get_tls()->tib_sigmask, -1);
}
static textwindows void UnblockSignals(uint64_t mask) {
atomic_store_explicit(&__get_tls()->tib_sigmask, mask, memory_order_release);
}
static textwindows int RaiseSignal(int sig) {
__get_tls()->tib_sigpending |= 1ull << (sig - 1);
return 0;
}
static textwindows int GetVirtualKey(uint16_t vk, bool shift, bool ctrl) {
for (int i = 0; i < ARRAYLEN(kVirtualKey); ++i) {
if (kVirtualKey[i].vk == vk) {
if (shift && ctrl) {
return kVirtualKey[i].shift_ctrl_str;
} else if (shift) {
return kVirtualKey[i].shift_str;
} else if (ctrl) {
return kVirtualKey[i].ctrl_str;
} else {
return kVirtualKey[i].normal_str;
}
}
}
return 0;
}
static textwindows int ProcessKeyEvent(const struct NtInputRecord *r, char *p) {
uint16_t c = r->Event.KeyEvent.uChar.UnicodeChar;
uint16_t vk = r->Event.KeyEvent.wVirtualKeyCode;
uint16_t cks = r->Event.KeyEvent.dwControlKeyState;
// ignore keyup events
if (!r->Event.KeyEvent.bKeyDown && (!c || vk != kNtVkMenu)) {
return 0;
}
// process arrow keys, function keys, etc.
int n = 0;
if (!c) {
int w;
w = GetVirtualKey(vk, !!(cks & kNtShiftPressed),
!!(cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed)));
if (!w) return 0;
p[n++] = 033;
if (cks & (kNtLeftAltPressed | kNtRightAltPressed)) {
p[n++] = 033;
}
if (w > 0) {
p[n++] = '[';
} else {
w = -w;
}
do p[n++] = w;
while ((w >>= 8));
return n;
}
// translate utf-16 into utf-32
if (IsHighSurrogate(c)) {
__keystroke.utf16hs = c;
return 0;
}
if (IsLowSurrogate(c)) {
c = MergeUtf16(__keystroke.utf16hs, c);
}
// enter sends \r in a raw terminals
// make it a multics newline instead
if (c == '\r' && !(__ttyconf.magic & kTtyNoCr2Nl)) {
c = '\n';
}
// microsoft doesn't encode ctrl-space (^@) as nul
// detecting it is also impossible w/ kNtEnableVirtualTerminalInput
if (c == ' ' && (cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
c = '\0';
}
// make it possible to distinguish ctrl-h (^H) from backspace (^?)
2023-10-04 06:53:59 +00:00
if (c == kNtVkBack && !(cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed))) {
c = 0177;
}
// handle ctrl-c and ctrl-\, which tcsetattr() is able to remap
if (c && !(__ttyconf.magic & kTtyNoIsigs)) {
if (c == __ttyconf.vintr) {
return RaiseSignal(SIGINT);
} else if (c == __ttyconf.vquit) {
return RaiseSignal(SIGQUIT);
}
}
// handle ctrl-d which generates end-of-file, unless pending line data
// is present, in which case we flush that without the newline instead
if (c && c == __ttyconf.veof && !(__ttyconf.magic & kTtyUncanon)) {
if (dll_is_empty(__keystroke.line)) {
2023-09-21 17:10:20 +00:00
__keystroke.end_of_file = true;
} else {
dll_make_last(&__keystroke.list, __keystroke.line);
__keystroke.line = 0;
2023-09-21 17:10:20 +00:00
}
return 0;
}
// insert esc prefix when alt is held
if ((cks & (kNtLeftAltPressed | kNtRightAltPressed)) &&
!(cks & (kNtLeftCtrlPressed | kNtRightCtrlPressed)) &&
r->Event.KeyEvent.bKeyDown) {
p[n++] = 033;
}
// convert utf-32 to utf-8
uint64_t w = tpenc(c);
do p[n++] = w;
while ((w >>= 8));
return n;
}
// To use the tty mouse events feature:
// - write(1, "\e[?1000;1002;1015;1006h") to enable
// - write(1, "\e[?1000;1002;1015;1006l") to disable
// See o//examples/ttyinfo.com and o//tool/viz/life.com
static textwindows int ProcessMouseEvent(const struct NtInputRecord *r,
char *b) {
char *p = b;
unsigned char e = 0;
uint32_t currentbs = __ttyconf.mousebs;
uint32_t ev = r->Event.MouseEvent.dwEventFlags;
uint32_t bs = r->Event.MouseEvent.dwButtonState;
ev &= kNtMouseMoved | kNtMouseWheeled;
bs &= kNtFromLeft1stButtonPressed | kNtRightmostButtonPressed;
if (ev & kNtMouseWheeled) {
// scroll wheel (unnatural mode)
if (!(r->Event.MouseEvent.dwControlKeyState &
(kNtShiftPressed | kNtLeftCtrlPressed | kNtRightCtrlPressed |
kNtLeftAltPressed | kNtRightAltPressed))) {
bool isup = ((int)r->Event.MouseEvent.dwButtonState >> 16) > 0;
if (__ttyconf.magic & kTtyXtMouse) {
e = isup ? 80 : 81;
goto OutputXtermMouseEvent;
} else {
// we disable mouse highlighting when the tty is put in raw mode
// to mouse wheel events with widely understood vt100 arrow keys
*p++ = 033;
*p++ = '[';
if (isup) {
*p++ = 'A'; // \e[A up
} else {
*p++ = 'B'; // \e[B down
}
}
}
} else if ((bs || currentbs) && (__ttyconf.magic & kTtyXtMouse)) {
if (bs && (ev & kNtMouseMoved) && currentbs) {
e |= 32; // dragging
}
if ((bs | currentbs) & kNtRightmostButtonPressed) {
e |= 2; // right
}
OutputXtermMouseEvent:
*p++ = 033;
*p++ = '[';
*p++ = '<';
p = FormatInt32(p, e);
*p++ = ';';
p = FormatInt32(p, (r->Event.MouseEvent.dwMousePosition.X + 1) & 0x7fff);
*p++ = ';';
p = FormatInt32(p, (r->Event.MouseEvent.dwMousePosition.Y + 1) & 0x7fff);
if (!bs && currentbs) {
*p++ = 'm'; // up
} else {
*p++ = 'M'; // down
}
__ttyconf.mousebs = bs;
}
return p - b;
}
static textwindows int ConvertConsoleInputToAnsi(const struct NtInputRecord *r,
char p[hasatleast 23]) {
switch (r->EventType) {
case kNtKeyEvent:
return ProcessKeyEvent(r, p);
case kNtMouseEvent:
return ProcessMouseEvent(r, p);
case kNtWindowBufferSizeEvent:
return RaiseSignal(SIGWINCH);
default:
return 0;
}
}
static textwindows struct Keystroke *NewKeystroke(void) {
struct Dll *e;
struct Keystroke *k = 0;
int i, n = ARRAYLEN(__keystroke.pool);
if (atomic_load_explicit(&__keystroke.pc, memory_order_acquire) < n &&
(i = atomic_fetch_add(&__keystroke.pc, 1)) < n) {
k = __keystroke.pool + i;
} else {
if ((e = dll_first(__keystroke.free))) {
k = KEYSTROKE_CONTAINER(e);
dll_remove(&__keystroke.free, &k->elem);
}
if (!k) {
if (_weaken(malloc)) {
k = _weaken(malloc)(sizeof(struct Keystroke));
} else {
enomem();
return 0;
}
}
}
if (k) {
bzero(k, sizeof(*k));
dll_init(&k->elem);
}
return k;
}
static textwindows void WriteTty(struct Fd *f, const char *p, size_t n) {
int64_t hOutput;
uint32_t dwConsoleMode;
if (f->kind == kFdConsole) {
hOutput = f->extra;
} else if (g_fds.p[1].kind == kFdFile &&
GetConsoleMode(g_fds.p[1].handle, &dwConsoleMode)) {
hOutput = g_fds.p[1].handle;
} else if (g_fds.p[2].kind == kFdFile &&
GetConsoleMode(g_fds.p[2].handle, &dwConsoleMode)) {
hOutput = g_fds.p[2].handle;
} else {
hOutput = g_fds.p[1].handle;
}
WriteFile(hOutput, p, n, 0, 0);
}
static textwindows bool IsCtl(int c) {
return isascii(c) && iscntrl(c) && c != '\n' && c != '\t';
}
static textwindows void WriteTtyCtl(struct Fd *f, const char *p, size_t n) {
size_t i;
for (i = 0; i < n; ++i) {
if (IsCtl(p[i])) {
char ctl[2];
ctl[0] = '^';
ctl[1] = p[i] ^ 0100;
WriteTty(f, ctl, 2);
} else {
WriteTty(f, p + i, 1);
}
}
}
static textwindows void EchoTty(struct Fd *f, const char *p, size_t n) {
if (__ttyconf.magic & kTtyEchoRaw) {
WriteTty(f, p, n);
} else {
WriteTtyCtl(f, p, n);
}
}
2023-10-04 06:53:59 +00:00
static textwindows bool EraseKeystroke(struct Fd *f) {
struct Dll *e;
if ((e = dll_last(__keystroke.line))) {
struct Keystroke *k = KEYSTROKE_CONTAINER(e);
dll_remove(&__keystroke.line, e);
dll_make_first(&__keystroke.free, e);
for (int i = k->buflen; i--;) {
if ((k->buf[i] & 0300) == 0200) continue;
WriteTty(f, "\b \b", 3);
if (!(__ttyconf.magic & kTtyEchoRaw) && IsCtl(k->buf[i])) {
WriteTty(f, "\b \b", 3);
}
}
return true;
} else {
return false;
}
}
static textwindows void IngestConsoleInputRecord(struct Fd *f,
struct NtInputRecord *r) {
// convert win32 console event into ansi
int len;
char buf[23];
if (!(len = ConvertConsoleInputToAnsi(r, buf))) {
return;
}
// handle backspace in canonical mode
if (len == 1 && buf[0] && //
(buf[0] & 255) == __ttyconf.verase && //
!(__ttyconf.magic & kTtyUncanon)) {
2023-10-04 06:53:59 +00:00
EraseKeystroke(f);
return;
}
// handle kill in canonical mode
if (len == 1 && buf[0] && //
(buf[0] & 255) == __ttyconf.vkill && //
!(__ttyconf.magic & kTtyUncanon)) {
while (EraseKeystroke(f)) {
}
return;
}
// allocate an object to hold this keystroke
struct Keystroke *k;
if (!(k = NewKeystroke())) {
STRACE("ran out of memory to hold keystroke %#.*s", len, buf);
return;
}
memcpy(k->buf, buf, sizeof(k->buf));
k->buflen = len;
// echo input if it was successfully recorded
// assuming the win32 console isn't doing it already
if (!(__ttyconf.magic & kTtySilence)) {
EchoTty(f, buf, len);
}
// save keystroke to appropriate list
if (__ttyconf.magic & kTtyUncanon) {
dll_make_last(&__keystroke.list, &k->elem);
} else {
dll_make_last(&__keystroke.line, &k->elem);
// handle end-of-line in canonical mode
if (len == 1 && buf[0] &&
((buf[0] & 255) == '\n' || //
(buf[0] & 255) == __ttyconf.veol || //
(buf[0] & 255) == __ttyconf.veol2)) {
dll_make_last(&__keystroke.list, __keystroke.line);
__keystroke.line = 0;
return;
}
}
}
static textwindows void IngestConsoleInput(struct Fd *f) {
uint32_t i, n;
struct NtInputRecord records[16];
if (!__keystroke.end_of_file) {
do {
2023-10-04 06:53:59 +00:00
if (GetNumberOfConsoleInputEvents(f->handle, &n)) {
if (n) {
n = MIN(ARRAYLEN(records), n);
if (ReadConsoleInput(f->handle, records, n, &n)) {
for (i = 0; i < n && !__keystroke.end_of_file; ++i) {
IngestConsoleInputRecord(f, records + i);
}
} else {
STRACE("ReadConsoleInput failed w/ %d", GetLastError());
__keystroke.end_of_file = true;
break;
}
}
} else {
2023-10-04 06:53:59 +00:00
STRACE("GetNumberOfConsoleInputRecords failed w/ %d", GetLastError());
__keystroke.end_of_file = true;
break;
}
} while (n == ARRAYLEN(records));
}
}
textwindows int FlushConsoleInputBytes(int64_t handle) {
int rc;
uint64_t m;
m = BlockSignals();
LockKeystrokes();
if (FlushConsoleInputBuffer(handle)) {
dll_make_first(&__keystroke.free, __keystroke.list);
__keystroke.list = 0;
2023-10-04 15:07:07 +00:00
dll_make_first(&__keystroke.free, __keystroke.line);
__keystroke.line = 0;
rc = 0;
} else {
rc = __winerr();
}
UnlockKeystrokes();
UnblockSignals(m);
return rc;
}
textwindows int CountConsoleInputBytes(struct Fd *f) {
int count = 0;
struct Dll *e;
uint64_t m = BlockSignals();
LockKeystrokes();
IngestConsoleInput(f);
for (e = dll_first(__keystroke.list); e; e = dll_next(__keystroke.list, e)) {
count += KEYSTROKE_CONTAINER(e)->buflen;
}
if (!count && __keystroke.end_of_file) {
count = -1;
}
UnlockKeystrokes();
UnblockSignals(m);
return count;
}
static textwindows bool DigestConsoleInput(char *data, size_t size, int *rc) {
// handle eof once available input is consumed
if (dll_is_empty(__keystroke.list) && __keystroke.end_of_file) {
*rc = 0;
return true;
}
// copy keystroke(s) into user buffer
int toto = 0;
struct Dll *e;
2023-10-04 06:53:59 +00:00
while (size && (e = dll_first(__keystroke.list))) {
struct Keystroke *k = KEYSTROKE_CONTAINER(e);
uint32_t got = MIN(size, k->buflen);
uint32_t remain = k->buflen - got;
if (got) {
memcpy(data, k->buf, got);
data += got;
size -= got;
toto += got;
}
if (remain) {
memmove(k->buf, k->buf + got, remain);
} else {
dll_remove(&__keystroke.list, e);
dll_make_first(&__keystroke.free, e);
}
k->buflen = remain;
if ((__ttyconf.magic & kTtyUncanon) && toto >= __ttyconf.vmin) {
break;
}
}
// return result
if (toto) {
*rc = toto;
return true;
} else {
return false;
}
}
static textwindows ssize_t ReadFromWindowsConsole(struct Fd *f, void *data,
size_t size) {
int rc = -1;
for (;;) {
bool done = false;
uint64_t m;
m = BlockSignals();
LockKeystrokes();
IngestConsoleInput(f);
done = DigestConsoleInput(data, size, &rc);
UnlockKeystrokes();
UnblockSignals(m);
if (done) break;
if (f->flags & O_NONBLOCK) return eagain();
uint32_t ms = __SIG_POLL_INTERVAL_MS;
if (!__ttyconf.vmin) {
if (!__ttyconf.vtime) {
return 0;
} else {
ms = __ttyconf.vtime * 100;
}
}
if (_check_interrupts(kSigOpRestartable)) return -1;
2023-10-04 06:53:59 +00:00
if (__pause_thread(ms)) {
if (errno == EAGAIN) {
errno = EINTR; // TODO(jart): Why do we need it?
}
return -1;
}
}
return rc;
}
textwindows ssize_t sys_read_nt_impl(int fd, void *data, size_t size,
int64_t offset) {
2023-07-23 17:57:18 +00:00
bool32 ok;
struct Fd *f;
2023-07-23 17:57:18 +00:00
uint32_t got;
int64_t handle;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
struct PosixThread *pt;
f = g_fds.p + fd;
handle = f->handle;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
pt = _pthread_self();
pt->abort_errno = EAGAIN;
size = MIN(size, 0x7ffff000);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
bool pwriting = offset != -1;
bool seekable = f->kind == kFdFile && GetFileType(handle) == kNtFileTypeDisk;
bool nonblock = !!(f->flags & O_NONBLOCK);
if (pwriting && !seekable) {
return espipe();
}
if (!pwriting) {
offset = 0;
}
uint32_t cm;
if (!seekable && (f->kind == kFdConsole || GetConsoleMode(handle, &cm))) {
return ReadFromWindowsConsole(f, data, size);
}
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!pwriting && seekable) {
pthread_mutex_lock(&f->lock);
offset = f->pointer;
}
struct NtOverlapped overlap = {.hEvent = CreateEvent(0, 0, 0, 0),
.Pointer = offset};
// the win32 manual says it's important to *not* put &got here
// since for overlapped i/o, we always use GetOverlappedResult
ok = ReadFile(handle, data, size, 0, &overlap);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!ok && GetLastError() == kNtErrorIoPending) {
BlockingOperation:
if (!nonblock) {
pt->ioverlap = &overlap;
pt->iohandle = handle;
}
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (nonblock) {
CancelIoEx(handle, &overlap);
} else if (_check_interrupts(kSigOpRestartable)) {
Interrupted:
pt->abort_errno = errno;
CancelIoEx(handle, &overlap);
} else {
for (;;) {
uint32_t i;
if (g_fds.stdin.inisem) {
ReleaseSemaphore(g_fds.stdin.inisem, 1, 0);
}
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
i = WaitForSingleObject(overlap.hEvent, __SIG_IO_INTERVAL_MS);
if (i == kNtWaitTimeout) {
if (_check_interrupts(kSigOpRestartable)) {
goto Interrupted;
}
} else {
break;
}
}
}
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
pt->ioverlap = 0;
pt->iohandle = 0;
ok = true;
}
if (ok) {
// overlapped is allocated on stack, so it's important we wait
// for windows to acknowledge that it's done using that memory
ok = GetOverlappedResult(handle, &overlap, &got, nonblock);
if (!ok && GetLastError() == kNtErrorIoIncomplete) {
goto BlockingOperation;
2023-07-23 17:57:18 +00:00
}
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
}
Make improvements - This change fixes a bug that allowed unbuffered printf() output (to streams like stderr) to be truncated. This regression was introduced some time between now and the last release. - POSIX specifies all functions as thread safe by default. This change works towards cleaning up our use of the @threadsafe / @threadunsafe documentation annotations to reflect that. The goal is (1) to use @threadunsafe to document functions which POSIX say needn't be thread safe, and (2) use @threadsafe to document functions that we chose to implement as thread safe even though POSIX didn't mandate it. - Tidy up the clock_gettime() implementation. We're now trying out a cleaner approach to system call support that aims to maintain the Linux errno convention as long as possible. This also fixes bugs that existed previously, where the vDSO errno wasn't being translated properly. The gettimeofday() system call is now a wrapper for clock_gettime(), which reduces bloat in apps that use both. - The recently-introduced improvements to the execute bit on Windows has had bugs fixed. access(X_OK) on a directory on Windows now succeeds. fstat() will now perform the MZ/#! ReadFile() operation correctly. - Windows.h is no longer included in libc/isystem/, because it confused PCRE's build system into thinking Cosmopolitan is a WIN32 platform. Cosmo's Windows.h polyfill was never even really that good, since it only defines a subset of the subset of WIN32 APIs that Cosmo defines. - The setlongerjmp() / longerjmp() APIs are removed. While they're nice APIs that are superior to the standardized setjmp / longjmp functions, they weren't superior enough to not be dead code in the monorepo. If you use these APIs, please file an issue and they'll be restored. - The .com appending magic has now been removed from APE Loader.
2023-10-03 02:25:19 +00:00
CloseHandle(overlap.hEvent);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!pwriting && seekable) {
if (ok) f->pointer = offset + got;
pthread_mutex_unlock(&f->lock);
}
if (ok) {
return got;
}
errno_t err;
if (_weaken(pthread_testcancel_np) &&
(err = _weaken(pthread_testcancel_np)())) {
return ecanceled();
}
switch (GetLastError()) {
case kNtErrorBrokenPipe: // broken pipe
case kNtErrorNoData: // closing named pipe
case kNtErrorHandleEof: // pread read past EOF
return 0; //
case kNtErrorAccessDenied: // read doesn't return EACCESS
return ebadf(); //
case kNtErrorOperationAborted:
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
errno = pt->abort_errno;
return -1;
default:
return __winerr();
}
}
textwindows ssize_t sys_read_nt(int fd, const struct iovec *iov, size_t iovlen,
int64_t opt_offset) {
ssize_t rc;
2021-02-07 14:11:44 +00:00
size_t i, total;
if (opt_offset < -1) return einval();
2020-06-15 14:18:57 +00:00
while (iovlen && !iov[0].iov_len) iov++, iovlen--;
2021-02-07 14:11:44 +00:00
if (iovlen) {
for (total = i = 0; i < iovlen; ++i) {
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
// TODO(jart): disable cancelations after first iteration
if (!iov[i].iov_len) continue;
rc = sys_read_nt_impl(fd, iov[i].iov_base, iov[i].iov_len, opt_offset);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (rc == -1) {
if (total && errno != ECANCELED) {
return total;
} else {
return -1;
}
}
total += rc;
if (opt_offset != -1) opt_offset += rc;
if (rc < iov[i].iov_len) break;
2021-02-07 14:11:44 +00:00
}
return total;
2020-06-15 14:18:57 +00:00
} else {
return sys_read_nt_impl(fd, NULL, 0, opt_offset);
2020-06-15 14:18:57 +00:00
}
}
#endif /* __x86_64__ */