cosmopolitan/libc/intrin/mman.greg.c

343 lines
15 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
2020-06-15 14:18:57 +00:00
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
αcτµαlly pδrταblε εxεcµταblε § no-frills virtual memory management
2020-06-15 14:18:57 +00:00
*/
#include "ape/relocations.h"
#include "libc/assert.h"
#include "libc/elf/def.h"
#include "libc/elf/struct/phdr.h"
#include "libc/macros.internal.h"
#include "libc/nexgen32e/uart.internal.h"
#include "libc/runtime/e820.internal.h"
#include "libc/runtime/metalprintf.internal.h"
#include "libc/runtime/pc.internal.h"
#include "libc/runtime/runtime.h"
2023-05-02 02:43:59 +00:00
#ifdef __x86_64__
2020-06-15 14:18:57 +00:00
#define INVERT(x) (BANE + PHYSICAL(x))
2023-04-27 03:45:01 +00:00
#define NOPAGE ((uint64_t)-1)
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
#define ABS64(x) \
({ \
int64_t vAddr; \
__asm__("movabs\t%1,%0" : "=r"(vAddr) : "i"(x)); \
vAddr; \
})
struct ReclaimedPage {
uint64_t next;
};
2020-06-15 14:18:57 +00:00
/**
* Allocates new page of physical memory.
2020-06-15 14:18:57 +00:00
*/
2023-09-12 08:07:51 +00:00
texthead uint64_t __new_page(struct mman *mm) {
uint64_t p = mm->frp;
if (p != NOPAGE) {
uint64_t q;
struct ReclaimedPage *rp = (struct ReclaimedPage *)(BANE + p);
unassert(p == (p & PAGE_TA));
q = rp->next;
unassert(q == (q & PAGE_TA) || q == NOPAGE);
mm->frp = q;
return p;
}
if (mm->pdpi == mm->e820n) {
return 0;
}
while (mm->pdp >= mm->e820[mm->pdpi].addr + mm->e820[mm->pdpi].size) {
if (++mm->pdpi == mm->e820n) return 0;
mm->pdp = MAX(mm->pdp, mm->e820[mm->pdpi].addr);
}
p = mm->pdp;
mm->pdp += 4096;
return p;
}
2020-06-15 14:18:57 +00:00
/**
* Returns pointer to page table entry for page at virtual address.
* Additional page tables are allocated if needed as a side-effect.
*/
2023-09-12 08:07:51 +00:00
textreal uint64_t *__get_virtual(struct mman *mm, uint64_t *t, int64_t vaddr,
bool maketables) {
uint64_t *e, p;
unsigned char h;
for (h = 39;; h -= 9) {
e = t + ((vaddr >> h) & 511);
if (h == 12) return e;
if (!(*e & (PAGE_V | PAGE_RSRV))) {
if (!maketables) return NULL;
if (!(p = __new_page(mm))) return NULL;
__clear_page(BANE + p);
*e = p | PAGE_V | PAGE_RW;
}
t = (uint64_t *)(BANE + (*e & PAGE_TA));
}
}
2020-06-15 14:18:57 +00:00
/**
* Sorts, rounds, and filters BIOS memory map.
*/
2023-09-12 08:07:51 +00:00
static textreal void __normalize_e820(struct mman *mm, uint64_t top) {
uint64_t a, b;
uint64_t x, y;
unsigned i, j, n;
for (n = i = 0; mm->e820[i].size; ++i) {
mm->e820[n] = mm->e820[i];
x = mm->e820[n].addr;
y = mm->e820[n].addr + mm->e820[n].size;
a = ROUNDUP(x, 4096);
b = ROUNDDOWN(y, 4096);
if (b > a && mm->e820[i].type == kMemoryUsable) {
b -= a;
mm->e820[n].addr = a;
mm->e820[n].size = b;
++n;
}
}
for (i = 1; i < n; ++i) {
for (j = i; j > 0 && mm->e820[i].addr < mm->e820[j - 1].addr; --j) {
mm->e820[j] = mm->e820[j - 1];
}
mm->e820[j] = mm->e820[i];
}
top = ROUNDUP(top, 4096);
mm->pdp = MAX(top, mm->e820[0].addr);
mm->pdpi = 0;
mm->e820n = n;
mm->frp = NOPAGE;
}
2020-06-15 14:18:57 +00:00
/**
* Identity maps an area of physical memory to its negative address.
*/
2023-09-12 08:07:51 +00:00
textreal uint64_t *__invert_memory_area(struct mman *mm, uint64_t *pml4t,
uint64_t ps, uint64_t size,
uint64_t pte_flags) {
uint64_t pe = ps + size, p, *m = NULL;
ps = ROUNDDOWN(ps, 4096);
pe = ROUNDUP(pe, 4096);
for (p = ps; p != pe; p += 4096) {
m = __get_virtual(mm, pml4t, BANE + p, true);
if (m && !(*m & (PAGE_V | PAGE_RSRV))) {
*m = p | PAGE_V | PAGE_RSRV | pte_flags;
}
}
return m;
}
/**
* Increments the reference count for a page of physical memory.
*/
2023-09-12 08:07:51 +00:00
void __ref_page(struct mman *mm, uint64_t *pml4t, uint64_t p) {
uint64_t *m, e;
m = __invert_memory_area(mm, pml4t, p, 4096, PAGE_RW | PAGE_XD);
if (m) {
e = *m;
if ((e & PAGE_REFC) != PAGE_REFC) {
e += PAGE_1REF;
*m = e;
}
}
}
/**
* Increments the reference counts for an area of physical memory.
*/
2023-09-12 08:07:51 +00:00
void __ref_pages(struct mman *mm, uint64_t *pml4t, uint64_t ps, uint64_t size) {
uint64_t p = ROUNDDOWN(ps, 4096), e = ROUNDUP(ps + size, 4096);
while (p != e) {
__ref_page(mm, pml4t, p);
p += 4096;
}
}
/**
* Reclaims a page of physical memory for later use.
*/
2023-09-12 08:07:51 +00:00
static void __reclaim_page(struct mman *mm, uint64_t p) {
struct ReclaimedPage *rp = (struct ReclaimedPage *)(BANE + p);
unassert(p == (p & PAGE_TA));
rp->next = mm->frp;
mm->frp = p;
}
/**
* Decrements the reference count for a page of physical memory. Frees the
* page if there are no virtual addresses (excluding the negative space)
* referring to it.
*/
2023-09-12 08:07:51 +00:00
void __unref_page(struct mman *mm, uint64_t *pml4t, uint64_t p) {
uint64_t *m, e;
m = __invert_memory_area(mm, pml4t, p, 4096, PAGE_RW | PAGE_XD);
if (m) {
e = *m;
if ((e & PAGE_REFC) != PAGE_REFC) {
e -= PAGE_1REF;
*m = e;
if ((e & PAGE_REFC) == 0) __reclaim_page(mm, p);
}
}
}
/**
* Identity maps all usable physical memory to its negative address.
*/
2023-09-12 08:07:51 +00:00
static textreal void __invert_memory(struct mman *mm, uint64_t *pml4t) {
uint64_t i;
for (i = 0; i < mm->e820n; ++i) {
uint64_t ps = mm->e820[i].addr, size = mm->e820[i].size;
/* ape/ape.S has already mapped the first 2 MiB of physical memory. */
if (ps < 0x200000 && ps + size <= 0x200000) continue;
__invert_memory_area(mm, pml4t, ps, size, PAGE_RW | PAGE_XD);
}
}
/**
* Exports information about the offset of a field within a structure type,
* so that assembly language routines can use it. This macro can be invoked
* from inside a function whose code is known to be emitted.
*/
#define export_offsetof(type, member) \
do { \
asm volatile(".globl \"" #type "::" #member "\"\n\t" \
".set \"" #type "::" #member "\",%c0" \
: /* no outputs */ \
: "i"(offsetof(type, member))); \
} while (0)
2023-09-12 08:07:51 +00:00
textreal void __setup_mman(struct mman *mm, uint64_t *pml4t, uint64_t top) {
export_offsetof(struct mman, pc_drive_base_table);
export_offsetof(struct mman, pc_drive_last_sector);
export_offsetof(struct mman, pc_drive_last_head);
export_offsetof(struct mman, pc_drive);
export_offsetof(struct mman, e820);
export_offsetof(struct mman, e820_end);
export_offsetof(struct mman, bad_idt);
export_offsetof(struct mman, pc_drive_next_sector);
export_offsetof(struct mman, pc_drive_next_cylinder);
export_offsetof(struct mman, pc_drive_next_head);
export_offsetof(struct mman, pc_video_type);
export_offsetof(struct mman, pc_video_stride);
export_offsetof(struct mman, pc_video_width);
export_offsetof(struct mman, pc_video_height);
export_offsetof(struct mman, pc_video_framebuffer);
export_offsetof(struct mman, pc_video_framebuffer_size);
export_offsetof(struct mman, pc_video_curs_info);
export_offsetof(struct mman, pc_video_char_height);
__normalize_e820(mm, top);
__invert_memory(mm, pml4t);
}
2020-06-15 14:18:57 +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
static textreal uint64_t __map_phdr(struct mman *mm, uint64_t *pml4t,
uint64_t b, uint64_t m,
struct Elf64_Phdr *p) {
uint64_t i, f, v;
f = PAGE_RSRV | PAGE_U;
if (p->p_flags & PF_W)
f |= PAGE_V | PAGE_RW;
else if (p->p_flags & (PF_R | PF_X))
f |= PAGE_V;
if (!(p->p_flags & PF_X)) f |= PAGE_XD;
for (i = 0; i < p->p_memsz; i += 4096) {
if (i < p->p_filesz) {
v = b + p->p_offset + i;
m = MAX(m, v);
} else {
v = __clear_page(BANE + __new_page(mm));
}
*__get_virtual(mm, pml4t, p->p_vaddr + i, true) = (v & PAGE_TA) | f;
__ref_page(mm, pml4t, v & PAGE_TA);
}
return m;
}
/**
* Maps APE-defined ELF program headers into memory and clears BSS.
*/
2023-09-12 08:07:51 +00:00
textreal void __map_phdrs(struct mman *mm, uint64_t *pml4t, uint64_t b,
uint64_t top) {
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
uint64_t m;
struct Elf64_Phdr *p;
extern char ape_phdrs[] __attribute__((__weak__));
extern char ape_phdrs_end[] __attribute__((__weak__));
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
extern char ape_stack_pf[] __attribute__((__weak__));
extern char ape_stack_offset[] __attribute__((__weak__));
extern char ape_stack_vaddr[] __attribute__((__weak__));
extern char ape_stack_filesz[] __attribute__((__weak__));
extern char ape_stack_memsz[] __attribute__((__weak__));
__setup_mman(mm, pml4t, top);
for (p = (struct Elf64_Phdr *)INVERT(ape_phdrs), m = 0;
p < (struct Elf64_Phdr *)INVERT(ape_phdrs_end); ++p) {
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
m = __map_phdr(mm, pml4t, b, m, p);
}
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
m = __map_phdr(mm, pml4t, b, m,
&(struct Elf64_Phdr){
.p_flags = (uintptr_t)ape_stack_pf,
.p_offset = (uintptr_t)ape_stack_offset,
.p_vaddr = ABS64(ape_stack_vaddr),
.p_filesz = (uintptr_t)ape_stack_filesz,
.p_memsz = (uintptr_t)ape_stack_memsz,
});
mm->pdp = MAX(mm->pdp, m);
}
/**
* Reclaims memory pages which were used at boot time but which can now be
* made available for the application.
*/
2023-09-12 08:07:51 +00:00
textreal void __reclaim_boot_pages(struct mman *mm, uint64_t skip_start,
uint64_t skip_end) {
uint64_t p = mm->frp, q = IMAGE_BASE_REAL, i, n = mm->e820n, b, e;
for (i = 0; i < n; ++i) {
b = mm->e820[i].addr;
if (b >= IMAGE_BASE_PHYSICAL) break;
e = MIN(IMAGE_BASE_PHYSICAL, b + mm->e820[i].size);
q = MAX(IMAGE_BASE_REAL, b);
while (q < e) {
struct ReclaimedPage *rp;
if (q == skip_start) {
q = skip_end;
if (q >= e) break;
}
rp = (struct ReclaimedPage *)(BANE + q);
rp->next = p;
p = q;
q += 4096;
}
}
mm->frp = p;
}
2023-05-02 02:43:59 +00:00
#endif /* __x86_64__ */