mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-07-01 08:48:29 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Make improvements

Browse source 
- 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.
This commit is contained in:
Justine Tunney 2023-09-18 20:44:45 -07:00
parent c4eb838516
commit ec480f5aa0
No known key found for this signature in database
GPG key ID: BE714B4575D6E328
638 changed files with 7925 additions and 8282 deletions
Download patch file Download diff file Expand all files Collapse all files

101
libc/runtime/clone.c
View file

@ -33,6 +33,8 @@
#include "libc/intrin/weaken.h"
#include "libc/limits.h"
#include "libc/macros.internal.h"
#include "libc/mem/alloca.h"
#include "libc/nt/enum/processcreationflags.h"
#include "libc/nt/runtime.h"
#include "libc/nt/signals.h"
#include "libc/nt/synchronization.h"
@ -40,10 +42,12 @@
#include "libc/nt/thunk/msabi.h"
#include "libc/runtime/internal.h"
#include "libc/runtime/runtime.h"
#include "libc/runtime/stack.h"
#include "libc/runtime/syslib.internal.h"
#include "libc/sock/internal.h"
#include "libc/stdalign.internal.h"
#include "libc/str/str.h"
#include "libc/sysv/consts/clone.h"
#include "libc/sysv/consts/futex.h"
#include "libc/sysv/consts/nr.h"
#include "libc/sysv/consts/nrlinux.h"
@ -85,6 +89,8 @@ struct CloneArgs {
void *arg;
};
int __stack_call(void *, int, long, long, int (*)(void *, int), void *);
static struct CloneArgs *AllocateCloneArgs(char *stk, size_t stksz) {
return (struct CloneArgs *)(((uintptr_t)(stk + stksz) -
sizeof(struct CloneArgs)) &
@ -100,29 +106,15 @@ __msabi extern typeof(TlsSetValue) *const __imp_TlsSetValue;
__msabi extern typeof(ExitThread) *const __imp_ExitThread;
__msabi extern typeof(WakeByAddressAll) *const __imp_WakeByAddressAll;
int WinThreadLaunch(void *arg, // rdi
int tid, // rsi
int (*func)(void *, int), // rdx
intptr_t rsp); // rcx
// we can't log this function because:
// 1. windows owns the backtrace pointer right now
// 2. ftrace unwinds rbp to determine depth
// 3. tid in tls for ftrace isn't set yet
// we can't use address sanitizer because:
// 1. __asan_handle_no_return wipes stack
// 2. windows owns the stack memory right now
// we need win32 raw imports because:
// 1. generated thunks are function logged
dontasan dontinstrument static textwindows wontreturn void //
WinThreadEntry(int rdi, // rcx
int rsi, // rdx
int rdx, // r8
struct CloneArgs *wt) { // r9
static textwindows dontinstrument wontreturn void //
WinThreadEntry(int rdi, // rcx
int rsi, // rdx
int rdx, // r8
struct CloneArgs *wt) { // r9
int rc;
if (wt->tls) __set_tls_win32(wt->tls);
*wt->ctid = wt->tid;
rc = WinThreadLaunch(wt->arg, wt->tid, wt->func, (intptr_t)wt);
rc = __stack_call(wt->arg, wt->tid, 0, 0, wt->func, wt);
// we can now clear ctid directly since we're no longer using our own
// stack memory, which can now be safely free'd by the parent thread.
*wt->ztid = 0;
@ -130,7 +122,7 @@ WinThreadEntry(int rdi, // rcx
// since we didn't indirect this function through NT2SYSV() it's not
// safe to simply return, and as such, we need ExitThread().
__imp_ExitThread(rc);
notpossible;
__builtin_unreachable();
}
static textwindows errno_t CloneWindows(int (*func)(void *, int), char *stk,
@ -144,8 +136,12 @@ static textwindows errno_t CloneWindows(int (*func)(void *, int), char *stk,
wt->func = func;
wt->arg = arg;
wt->tls = flags & CLONE_SETTLS ? tls : 0;
if ((h = CreateThread(0, 65536, (void *)WinThreadEntry, wt, 0, &wt->utid))) {
CloseHandle(h);
if ((h = CreateThread(0, 65536, (void *)WinThreadEntry, wt,
kNtStackSizeParamIsAReservation, &wt->utid))) {
if (flags & CLONE_SETTLS) {
struct CosmoTib *tib = tls;
tib->tib_syshand = h;
}
if (flags & CLONE_PARENT_SETTID) {
*ptid = wt->tid;
}
@ -196,9 +192,6 @@ XnuThreadMain(void *pthread, // rdi
func(arg, tid);
// avoid signal handler being triggered after we trash our stack
_sigblockall();
// we no longer use the stack after this point
// %rax = int bsdthread_terminate(%rdi = void *stackaddr,
// %rsi = size_t freesize,
@ -210,7 +203,7 @@ XnuThreadMain(void *pthread, // rdi
: "=m"(*wt->ztid)
: "a"(0x2000000 | 361), "D"(0), "S"(0), "d"(0L)
: "rcx", "r10", "r11", "memory");
notpossible;
__builtin_unreachable();
}
static errno_t CloneXnu(int (*fn)(void *), char *stk, size_t stksz, int flags,
@ -236,8 +229,6 @@ static wontreturn void FreebsdThreadMain(void *p) {
struct CloneArgs *wt = p;
*wt->ctid = wt->tid;
wt->func(wt->arg, wt->tid);
// avoid signal handler being triggered after we trash our stack
_sigblockall();
// we no longer use the stack after this point
// void thr_exit(%rdi = long *state);
asm volatile("movl\t$0,%0\n\t" // *wt->ztid = 0
@ -250,7 +241,7 @@ static wontreturn void FreebsdThreadMain(void *p) {
: "=m"(*wt->ztid)
: "a"(454), "D"(wt->ztid), "S"(UMTX_OP_WAKE), "d"(INT_MAX)
: "rcx", "r8", "r9", "r10", "r11", "memory");
notpossible;
__builtin_unreachable();
}
static errno_t CloneFreebsd(int (*func)(void *, int), char *stk, size_t stksz,
@ -290,7 +281,7 @@ static errno_t CloneFreebsd(int (*func)(void *, int), char *stk, size_t stksz,
// we can't use address sanitizer because:
// 1. __asan_handle_no_return wipes stack [todo?]
dontasan static wontreturn void OpenbsdThreadMain(void *p) {
static wontreturn void OpenbsdThreadMain(void *p) {
struct CloneArgs *wt = p;
*wt->ctid = wt->tid;
wt->func(wt->arg, wt->tid);
@ -304,7 +295,7 @@ dontasan static wontreturn void OpenbsdThreadMain(void *p) {
: "a"(83), "m"(__oldstack), "D"(wt->ztid),
"S"(2 /* FUTEX_WAKE */), "d"(INT_MAX)
: "rcx", "r11", "memory");
notpossible;
__builtin_unreachable();
}
static errno_t CloneOpenbsd(int (*func)(void *, int), char *stk, size_t stksz,
@ -353,8 +344,6 @@ static wontreturn void NetbsdThreadMain(void *arg, // rdi
ax = sys_gettid();
*ctid = ax;
func(arg, ax);
// avoid signal handler being triggered after we trash our stack
_sigblockall();
// we no longer use the stack after this point
// %eax = int __lwp_exit(void);
asm volatile("movl\t$0,%2\n\t" // *wt->ztid = 0
@ -362,7 +351,7 @@ static wontreturn void NetbsdThreadMain(void *arg, // rdi
: "=a"(ax), "=d"(dx), "=m"(*ztid)
: "0"(310)
: "rcx", "r11", "memory");
notpossible;
__builtin_unreachable();
}
static int CloneNetbsd(int (*func)(void *, int), char *stk, size_t stksz,
@ -449,21 +438,10 @@ static int CloneNetbsd(int (*func)(void *, int), char *stk, size_t stksz,
// APPLE SILICON
static void *SiliconThreadMain(void *arg) {
register struct CloneArgs *wt asm("x21") = arg;
struct CloneArgs *wt = arg;
asm volatile("mov\tx28,%0" : /* no outputs */ : "r"(wt->tls));
*wt->ctid = wt->this;
register void *x0 asm("x0") = wt->arg;
register int x1 asm("w1") = wt->this;
register void *x28 asm("x28") = wt->tls;
asm volatile("mov\tx19,x29\n\t" // save frame pointer
"mov\tx20,sp\n\t" // save stack pointer
"mov\tx29,#0\n\t" // reset backtrace
"mov\tsp,%3\n\t" // switch stack
"blr\t%2\n\t" // wt->func(wt->arg, tid)
"mov\tx29,x19\n\t" // restore frame pointer
"mov\tsp,x20" // restore stack pointer
: "+r"(x0)
: "r"(x1), "r"(wt->func), "r"(wt), "r"(x28)
: "x19", "x20", "memory");
__stack_call(wt->arg, wt->this, 0, 0, wt->func, wt);
*wt->ztid = 0;
return 0;
}
@ -471,9 +449,11 @@ static void *SiliconThreadMain(void *arg) {
static errno_t CloneSilicon(int (*fn)(void *, int), char *stk, size_t stksz,
int flags, void *arg, void *tls, int *ptid,
int *ctid) {
void *attr;
errno_t res;
unsigned tid;
pthread_t th;
size_t babystack;
struct CloneArgs *wt;
static atomic_uint tids;
wt = AllocateCloneArgs(stk, stksz);
@ -484,14 +464,23 @@ static errno_t CloneSilicon(int (*fn)(void *, int), char *stk, size_t stksz,
wt->tls = flags & CLONE_SETTLS ? tls : 0;
wt->func = fn;
wt->arg = arg;
if (!(res = __syslib->pthread_create(&th, 0, SiliconThreadMain, wt)) &&
babystack = __syslib->__pthread_stack_min;
#pragma GCC push_options
#pragma GCC diagnostic ignored "-Walloca-larger-than="
attr = alloca(__syslib->__sizeof_pthread_attr_t);
#pragma GCC pop_options
unassert(!__syslib->__pthread_attr_init(attr));
unassert(!__syslib->__pthread_attr_setguardsize(attr, 0));
unassert(!__syslib->__pthread_attr_setstacksize(attr, babystack));
if (!(res = __syslib->__pthread_create(&th, attr, SiliconThreadMain, wt)) &&
(flags & CLONE_PARENT_SETTID)) {
*ptid = tid;
if (flags & CLONE_SETTLS) {
struct CosmoTib *tib = tls;
tib[-1].tib_pthread = th;
tib[-1].tib_syshand = th;
}
}
unassert(!__syslib->__pthread_attr_destroy(attr));
return res;
}
@ -620,9 +609,10 @@ static int CloneLinux(int (*func)(void *arg, int rc), char *stk, size_t stksz,
* child has started.
*
* - `CLONE_PARENT_SETTID` must be specified if you intend to set
* the `ptid` argument, which is updated at the most opportune
* moment. On all platforms except XNU, this happens before
* clone() returns. On XNU, it happens once the thread starts.
* the `ptid` argument, and it is updated at the most opportune
* moment. On all platforms except XNU x86, this happens before
* clone() returns. But since it might not be available yet you
* need to use pthread_getunique_np() to obtain it.
*
* - `CLONE_CHILD_CLEARTID` causes `*ctid = 0` upon child thread
* termination. This is used to implement join so that the parent
@ -653,9 +643,6 @@ errno_t clone(void *func, void *stk, size_t stksz, int flags, void *arg,
if (!func) {
rc = EINVAL;
} else if (!IsTiny() &&
((flags & CLONE_VM) && (stksz < 4096 || (stksz & 15)))) {
rc = EINVAL;
} else if (IsAsan() &&
(((flags & CLONE_SETTLS) && !__asan_is_valid(tls, 64)) ||
((flags & CLONE_PARENT_SETTID) &&