2021-07-19 21:55:20 +00:00
|
|
|
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
|
|
|
|
│ vi: set et ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi │
|
|
|
|
╞══════════════════════════════════════════════════════════════════════════════╡
|
|
|
|
│ Copyright The Mbed TLS Contributors │
|
|
|
|
│ │
|
|
|
|
│ Licensed under the Apache License, Version 2.0 (the "License"); │
|
|
|
|
│ you may not use this file except in compliance with the License. │
|
|
|
|
│ You may obtain a copy of the License at │
|
|
|
|
│ │
|
|
|
|
│ http://www.apache.org/licenses/LICENSE-2.0 │
|
|
|
|
│ │
|
|
|
|
│ Unless required by applicable law or agreed to in writing, software │
|
|
|
|
│ distributed under the License is distributed on an "AS IS" BASIS, │
|
|
|
|
│ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. │
|
|
|
|
│ See the License for the specific language governing permissions and │
|
|
|
|
│ limitations under the License. │
|
|
|
|
╚─────────────────────────────────────────────────────────────────────────────*/
|
2023-08-08 03:22:49 +00:00
|
|
|
#include "third_party/mbedtls/net_sockets.h"
|
2021-07-19 21:55:20 +00:00
|
|
|
#include "libc/calls/calls.h"
|
2021-10-14 00:27:13 +00:00
|
|
|
#include "libc/calls/struct/sigaction.h"
|
2023-08-08 03:22:49 +00:00
|
|
|
#include "libc/calls/weirdtypes.h"
|
2021-07-19 21:55:20 +00:00
|
|
|
#include "libc/errno.h"
|
|
|
|
#include "libc/sock/select.h"
|
2022-08-06 16:56:17 +00:00
|
|
|
#include "libc/sock/struct/sockaddr6.h"
|
2021-07-19 21:55:20 +00:00
|
|
|
#include "libc/sysv/consts/af.h"
|
|
|
|
#include "libc/sysv/consts/f.h"
|
|
|
|
#include "libc/sysv/consts/ipproto.h"
|
|
|
|
#include "libc/sysv/consts/msg.h"
|
|
|
|
#include "libc/sysv/consts/o.h"
|
2021-10-14 00:27:13 +00:00
|
|
|
#include "libc/sysv/consts/sig.h"
|
2021-07-19 21:55:20 +00:00
|
|
|
#include "libc/sysv/consts/so.h"
|
|
|
|
#include "libc/sysv/consts/sock.h"
|
|
|
|
#include "libc/sysv/consts/sol.h"
|
|
|
|
#include "third_party/mbedtls/error.h"
|
2023-12-29 06:58:17 +00:00
|
|
|
#include "third_party/musl/netdb.h"
|
|
|
|
#include "libc/sock/sock.h"
|
2021-07-19 21:55:20 +00:00
|
|
|
#include "third_party/mbedtls/ssl.h"
|
|
|
|
|
|
|
|
#define IS_EINTR(ret) ((ret) == EINTR)
|
|
|
|
|
|
|
|
static int net_prepare(void) {
|
|
|
|
signal(SIGPIPE, SIG_IGN);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Initialize a context
|
|
|
|
* Just makes the context ready to be used or freed safely.
|
|
|
|
*
|
|
|
|
* \param ctx Context to initialize
|
|
|
|
*/
|
|
|
|
void mbedtls_net_init(mbedtls_net_context *ctx) {
|
|
|
|
ctx->fd = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Initiate a connection with host:port in the given protocol
|
|
|
|
*
|
|
|
|
* \param ctx Socket to use
|
|
|
|
* \param host Host to connect to
|
|
|
|
* \param port Port to connect to
|
|
|
|
* \param proto Protocol: MBEDTLS_NET_PROTO_TCP or MBEDTLS_NET_PROTO_UDP
|
|
|
|
*
|
|
|
|
* \return 0 if successful, or one of:
|
|
|
|
* MBEDTLS_ERR_NET_SOCKET_FAILED,
|
|
|
|
* MBEDTLS_ERR_NET_UNKNOWN_HOST,
|
|
|
|
* MBEDTLS_ERR_NET_CONNECT_FAILED
|
|
|
|
*
|
|
|
|
* \note Sets the socket in connected mode even with UDP.
|
|
|
|
*/
|
|
|
|
int mbedtls_net_connect(mbedtls_net_context *ctx, const char *host,
|
|
|
|
const char *port, int proto) {
|
|
|
|
int ret = MBEDTLS_ERR_THIS_CORRUPTION;
|
|
|
|
struct addrinfo hints, *addr_list, *cur;
|
|
|
|
if ((ret = net_prepare()) != 0) return ret;
|
|
|
|
/* Do name resolution with both IPv6 and IPv4 */
|
|
|
|
mbedtls_platform_zeroize(&hints, sizeof(hints));
|
|
|
|
hints.ai_family = AF_UNSPEC;
|
|
|
|
hints.ai_socktype = proto == MBEDTLS_NET_PROTO_UDP ? SOCK_DGRAM : SOCK_STREAM;
|
|
|
|
hints.ai_protocol =
|
|
|
|
proto == MBEDTLS_NET_PROTO_UDP ? IPPROTO_UDP : IPPROTO_TCP;
|
|
|
|
if (getaddrinfo(host, port, &hints, &addr_list) != 0)
|
|
|
|
return MBEDTLS_ERR_NET_UNKNOWN_HOST;
|
|
|
|
/* Try the sockaddrs until a connection succeeds */
|
|
|
|
ret = MBEDTLS_ERR_NET_UNKNOWN_HOST;
|
|
|
|
for (cur = addr_list; cur != NULL; cur = cur->ai_next) {
|
|
|
|
ctx->fd = (int)socket(cur->ai_family, cur->ai_socktype, cur->ai_protocol);
|
|
|
|
if (ctx->fd < 0) {
|
|
|
|
ret = MBEDTLS_ERR_NET_SOCKET_FAILED;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (connect(ctx->fd, cur->ai_addr, cur->ai_addrlen) == 0) {
|
|
|
|
ret = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
close(ctx->fd);
|
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 (errno == ECANCELED) {
|
|
|
|
ret = MBEDTLS_ERR_SSL_CANCELED;
|
|
|
|
} else {
|
|
|
|
ret = MBEDTLS_ERR_NET_CONNECT_FAILED;
|
|
|
|
}
|
2021-07-19 21:55:20 +00:00
|
|
|
}
|
|
|
|
freeaddrinfo(addr_list);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Create a receiving socket on bind_ip:port in the chosen
|
|
|
|
* protocol. If bind_ip == NULL, all interfaces are bound.
|
|
|
|
*
|
|
|
|
* \param ctx Socket to use
|
|
|
|
* \param bind_ip IP to bind to, can be NULL
|
|
|
|
* \param port Port number to use
|
|
|
|
* \param proto Protocol: MBEDTLS_NET_PROTO_TCP or MBEDTLS_NET_PROTO_UDP
|
|
|
|
*
|
|
|
|
* \return 0 if successful, or one of:
|
|
|
|
* MBEDTLS_ERR_NET_SOCKET_FAILED,
|
|
|
|
* MBEDTLS_ERR_NET_UNKNOWN_HOST,
|
|
|
|
* MBEDTLS_ERR_NET_BIND_FAILED,
|
|
|
|
* MBEDTLS_ERR_NET_LISTEN_FAILED
|
|
|
|
*
|
|
|
|
* \note Regardless of the protocol, opens the sockets and binds it.
|
|
|
|
* In addition, make the socket listening if protocol is TCP.
|
|
|
|
*/
|
|
|
|
int mbedtls_net_bind(mbedtls_net_context *ctx, const char *bind_ip,
|
|
|
|
const char *port, int proto) {
|
|
|
|
int n, ret;
|
|
|
|
struct addrinfo hints, *addr_list, *cur;
|
|
|
|
if ((ret = net_prepare()) != 0) return ret;
|
|
|
|
/* Bind to IPv6 and/or IPv4, but only in the desired protocol */
|
|
|
|
mbedtls_platform_zeroize(&hints, sizeof(hints));
|
|
|
|
hints.ai_family = AF_UNSPEC;
|
|
|
|
hints.ai_socktype = proto == MBEDTLS_NET_PROTO_UDP ? SOCK_DGRAM : SOCK_STREAM;
|
|
|
|
hints.ai_protocol =
|
|
|
|
proto == MBEDTLS_NET_PROTO_UDP ? IPPROTO_UDP : IPPROTO_TCP;
|
|
|
|
if (bind_ip == NULL) hints.ai_flags = AI_PASSIVE;
|
|
|
|
if (getaddrinfo(bind_ip, port, &hints, &addr_list) != 0)
|
|
|
|
return MBEDTLS_ERR_NET_UNKNOWN_HOST;
|
|
|
|
/* Try the sockaddrs until a binding succeeds */
|
|
|
|
ret = MBEDTLS_ERR_NET_UNKNOWN_HOST;
|
|
|
|
for (cur = addr_list; cur != NULL; cur = cur->ai_next) {
|
|
|
|
ctx->fd = (int)socket(cur->ai_family, cur->ai_socktype, cur->ai_protocol);
|
|
|
|
if (ctx->fd < 0) {
|
|
|
|
ret = MBEDTLS_ERR_NET_SOCKET_FAILED;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
n = 1;
|
|
|
|
if (setsockopt(ctx->fd, SOL_SOCKET, SO_REUSEADDR, (const char *)&n,
|
|
|
|
sizeof(n)) != 0) {
|
|
|
|
close(ctx->fd);
|
|
|
|
ret = MBEDTLS_ERR_NET_SOCKET_FAILED;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (bind(ctx->fd, cur->ai_addr, cur->ai_addrlen) != 0) {
|
|
|
|
close(ctx->fd);
|
|
|
|
ret = MBEDTLS_ERR_NET_BIND_FAILED;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
/* Listen only makes sense for TCP */
|
|
|
|
if (proto == MBEDTLS_NET_PROTO_TCP) {
|
|
|
|
if (listen(ctx->fd, MBEDTLS_NET_LISTEN_BACKLOG) != 0) {
|
|
|
|
close(ctx->fd);
|
|
|
|
ret = MBEDTLS_ERR_NET_LISTEN_FAILED;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* Bind was successful */
|
|
|
|
ret = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
freeaddrinfo(addr_list);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if the requested operation would be blocking on a non-blocking socket
|
|
|
|
* and thus 'failed' with a negative return value.
|
|
|
|
*
|
|
|
|
* Note: on a blocking socket this function always returns 0!
|
|
|
|
*/
|
|
|
|
static int net_would_block(const mbedtls_net_context *ctx) {
|
|
|
|
int err = errno;
|
|
|
|
/*
|
|
|
|
* Never return 'WOULD BLOCK' on a blocking socket
|
|
|
|
*/
|
|
|
|
if ((fcntl(ctx->fd, F_GETFL) & O_NONBLOCK) != O_NONBLOCK) {
|
|
|
|
errno = err;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
errno = err;
|
|
|
|
if (err == EAGAIN || err == EWOULDBLOCK) return 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Accept a connection from a remote client
|
|
|
|
*
|
|
|
|
* \param bind_ctx Relevant socket
|
|
|
|
* \param client_ctx Will contain the connected client socket
|
|
|
|
* \param client_ip Will contain the client IP address, can be NULL
|
|
|
|
* \param buf_size Size of the client_ip buffer
|
|
|
|
* \param ip_len Will receive the size of the client IP written,
|
|
|
|
* can be NULL if client_ip is null
|
|
|
|
*
|
|
|
|
* \return 0 if successful, or
|
|
|
|
* MBEDTLS_ERR_NET_SOCKET_FAILED,
|
|
|
|
* MBEDTLS_ERR_NET_BIND_FAILED,
|
|
|
|
* MBEDTLS_ERR_NET_ACCEPT_FAILED, or
|
|
|
|
* MBEDTLS_ERR_NET_BUFFER_TOO_SMALL if buf_size is too small,
|
|
|
|
* MBEDTLS_ERR_SSL_WANT_READ if bind_fd was set to
|
|
|
|
* non-blocking and accept() would block.
|
|
|
|
*/
|
|
|
|
int mbedtls_net_accept(mbedtls_net_context *bind_ctx,
|
|
|
|
mbedtls_net_context *client_ctx, void *client_ip,
|
|
|
|
size_t buf_size, size_t *ip_len) {
|
|
|
|
int ret = MBEDTLS_ERR_THIS_CORRUPTION;
|
|
|
|
int type;
|
|
|
|
struct sockaddr_storage client_addr;
|
|
|
|
#if defined(__socklen_t_defined) || defined(_SOCKLEN_T) || \
|
|
|
|
defined(_SOCKLEN_T_DECLARED) || defined(__DEFINED_socklen_t) || \
|
|
|
|
defined(socklen_t) || \
|
|
|
|
(defined(_POSIX_VERSION) && _POSIX_VERSION >= 200112L)
|
|
|
|
socklen_t n = (socklen_t)sizeof(client_addr);
|
|
|
|
socklen_t type_len = (socklen_t)sizeof(type);
|
|
|
|
#else
|
|
|
|
int n = (int)sizeof(client_addr);
|
|
|
|
int type_len = (int)sizeof(type);
|
|
|
|
#endif
|
|
|
|
/* Is this a TCP or UDP socket? */
|
|
|
|
if (getsockopt(bind_ctx->fd, SOL_SOCKET, SO_TYPE, (void *)&type, &type_len) !=
|
|
|
|
0 ||
|
|
|
|
(type != SOCK_STREAM && type != SOCK_DGRAM)) {
|
|
|
|
return MBEDTLS_ERR_NET_ACCEPT_FAILED;
|
|
|
|
}
|
|
|
|
if (type == SOCK_STREAM) {
|
|
|
|
/* TCP: actual accept() */
|
|
|
|
ret = client_ctx->fd =
|
|
|
|
(int)accept(bind_ctx->fd, (struct sockaddr *)&client_addr, &n);
|
|
|
|
} else {
|
|
|
|
/* UDP: wait for a message, but keep it in the queue */
|
|
|
|
char buf[1] = {0};
|
|
|
|
ret = (int)recvfrom(bind_ctx->fd, buf, sizeof(buf), MSG_PEEK,
|
|
|
|
(struct sockaddr *)&client_addr, &n);
|
|
|
|
#if defined(_WIN32)
|
|
|
|
if (ret == SOCKET_ERROR && WSAGetLastError() == WSAEMSGSIZE) {
|
|
|
|
/* We know buf is too small, thanks, just peeking here */
|
|
|
|
ret = 0;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
if (ret < 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
|
|
|
if (errno == ECANCELED) return MBEDTLS_ERR_SSL_CANCELED;
|
2021-07-19 21:55:20 +00:00
|
|
|
if (net_would_block(bind_ctx) != 0) return MBEDTLS_ERR_SSL_WANT_READ;
|
|
|
|
return MBEDTLS_ERR_NET_ACCEPT_FAILED;
|
|
|
|
}
|
|
|
|
/* UDP: hijack the listening socket to communicate with the client,
|
|
|
|
* then bind a new socket to accept new connections */
|
|
|
|
if (type != SOCK_STREAM) {
|
|
|
|
struct sockaddr_storage local_addr;
|
|
|
|
int one = 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
|
|
|
if (connect(bind_ctx->fd, (struct sockaddr *)&client_addr, n) != 0) {
|
|
|
|
if (errno == ECANCELED) return MBEDTLS_ERR_SSL_CANCELED;
|
2021-07-19 21:55:20 +00:00
|
|
|
return MBEDTLS_ERR_NET_ACCEPT_FAILED;
|
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
|
|
|
}
|
2021-07-19 21:55:20 +00:00
|
|
|
client_ctx->fd = bind_ctx->fd;
|
|
|
|
bind_ctx->fd = -1; /* In case we exit early */
|
|
|
|
n = sizeof(struct sockaddr_storage);
|
|
|
|
if (getsockname(client_ctx->fd, (struct sockaddr *)&local_addr, &n) != 0 ||
|
|
|
|
(bind_ctx->fd =
|
|
|
|
(int)socket(local_addr.ss_family, SOCK_DGRAM, IPPROTO_UDP)) < 0 ||
|
|
|
|
setsockopt(bind_ctx->fd, SOL_SOCKET, SO_REUSEADDR, (const char *)&one,
|
|
|
|
sizeof(one)) != 0) {
|
|
|
|
return MBEDTLS_ERR_NET_SOCKET_FAILED;
|
|
|
|
}
|
|
|
|
if (bind(bind_ctx->fd, (struct sockaddr *)&local_addr, n) != 0) {
|
|
|
|
return MBEDTLS_ERR_NET_BIND_FAILED;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (client_ip != NULL) {
|
|
|
|
if (client_addr.ss_family == AF_INET) {
|
|
|
|
struct sockaddr_in *addr4 = (struct sockaddr_in *)&client_addr;
|
|
|
|
*ip_len = sizeof(addr4->sin_addr.s_addr);
|
|
|
|
if (buf_size < *ip_len) return MBEDTLS_ERR_NET_BUFFER_TOO_SMALL;
|
|
|
|
memcpy(client_ip, &addr4->sin_addr.s_addr, *ip_len);
|
|
|
|
} else {
|
|
|
|
struct sockaddr_in6 *addr6 = (struct sockaddr_in6 *)&client_addr;
|
|
|
|
*ip_len = sizeof(addr6->sin6_addr.s6_addr);
|
|
|
|
if (buf_size < *ip_len) return MBEDTLS_ERR_NET_BUFFER_TOO_SMALL;
|
|
|
|
memcpy(client_ip, &addr6->sin6_addr.s6_addr, *ip_len);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Set the socket blocking
|
|
|
|
*
|
|
|
|
* \param ctx Socket to set
|
|
|
|
*
|
|
|
|
* \return 0 if successful, or a non-zero error code
|
|
|
|
*/
|
|
|
|
int mbedtls_net_set_block(mbedtls_net_context *ctx) {
|
|
|
|
return fcntl(ctx->fd, F_SETFL, fcntl(ctx->fd, F_GETFL) & ~O_NONBLOCK);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Set the socket non-blocking
|
|
|
|
*
|
|
|
|
* \param ctx Socket to set
|
|
|
|
*
|
|
|
|
* \return 0 if successful, or a non-zero error code
|
|
|
|
*/
|
|
|
|
int mbedtls_net_set_nonblock(mbedtls_net_context *ctx) {
|
|
|
|
return fcntl(ctx->fd, F_SETFL, fcntl(ctx->fd, F_GETFL) | O_NONBLOCK);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Check and wait for the context to be ready for read/write
|
|
|
|
*
|
|
|
|
* \note The current implementation of this function uses
|
|
|
|
* select() and returns an error if the file descriptor
|
|
|
|
* is \c FD_SETSIZE or greater.
|
|
|
|
*
|
|
|
|
* \param ctx Socket to check
|
|
|
|
* \param rw Bitflag composed of MBEDTLS_NET_POLL_READ and
|
|
|
|
* MBEDTLS_NET_POLL_WRITE specifying the events
|
|
|
|
* to wait for:
|
|
|
|
* - If MBEDTLS_NET_POLL_READ is set, the function
|
|
|
|
* will return as soon as the net context is available
|
|
|
|
* for reading.
|
|
|
|
* - If MBEDTLS_NET_POLL_WRITE is set, the function
|
|
|
|
* will return as soon as the net context is available
|
|
|
|
* for writing.
|
|
|
|
* \param timeout Maximal amount of time to wait before returning,
|
|
|
|
* in milliseconds. If \c timeout is zero, the
|
|
|
|
* function returns immediately. If \c timeout is
|
|
|
|
* -1u, the function blocks potentially indefinitely.
|
|
|
|
*
|
|
|
|
* \return Bitmask composed of MBEDTLS_NET_POLL_READ/WRITE
|
|
|
|
* on success or timeout, or a negative return code otherwise.
|
|
|
|
*/
|
|
|
|
int mbedtls_net_poll(mbedtls_net_context *ctx, uint32_t rw, uint32_t timeout) {
|
|
|
|
int ret = MBEDTLS_ERR_THIS_CORRUPTION;
|
|
|
|
struct timeval tv;
|
|
|
|
fd_set read_fds;
|
|
|
|
fd_set write_fds;
|
|
|
|
int fd = ctx->fd;
|
|
|
|
if (fd < 0) return MBEDTLS_ERR_NET_INVALID_CONTEXT;
|
|
|
|
/* A limitation of select() is that it only works with file descriptors
|
|
|
|
* that are strictly less than FD_SETSIZE. This is a limitation of the
|
|
|
|
* fd_set type. Error out early, because attempting to call FD_SET on a
|
|
|
|
* large file descriptor is a buffer overflow on typical platforms. */
|
|
|
|
if (fd >= FD_SETSIZE) return MBEDTLS_ERR_NET_POLL_FAILED;
|
|
|
|
#if defined(__has_feature)
|
|
|
|
#if __has_feature(memory_sanitizer)
|
|
|
|
/* Ensure that memory sanitizers consider read_fds and write_fds as
|
|
|
|
* initialized even on platforms such as Glibc/x86_64 where FD_ZERO
|
|
|
|
* is implemented in assembly. */
|
|
|
|
mbedtls_platform_zeroize(&read_fds, sizeof(read_fds));
|
|
|
|
mbedtls_platform_zeroize(&write_fds, sizeof(write_fds));
|
|
|
|
#endif
|
|
|
|
#endif
|
|
|
|
FD_ZERO(&read_fds);
|
|
|
|
if (rw & MBEDTLS_NET_POLL_READ) {
|
|
|
|
rw &= ~MBEDTLS_NET_POLL_READ;
|
|
|
|
FD_SET(fd, &read_fds);
|
|
|
|
}
|
|
|
|
FD_ZERO(&write_fds);
|
|
|
|
if (rw & MBEDTLS_NET_POLL_WRITE) {
|
|
|
|
rw &= ~MBEDTLS_NET_POLL_WRITE;
|
|
|
|
FD_SET(fd, &write_fds);
|
|
|
|
}
|
|
|
|
if (rw != 0) return MBEDTLS_ERR_NET_BAD_INPUT_DATA;
|
|
|
|
tv.tv_sec = timeout / 1000;
|
|
|
|
tv.tv_usec = (timeout % 1000) * 1000;
|
|
|
|
do {
|
|
|
|
ret = select(fd + 1, &read_fds, &write_fds, NULL,
|
|
|
|
timeout == (uint32_t)-1 ? NULL : &tv);
|
|
|
|
} while (IS_EINTR(ret));
|
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 (ret < 0) {
|
|
|
|
if (errno == ECANCELED) return MBEDTLS_ERR_SSL_CANCELED;
|
|
|
|
return MBEDTLS_ERR_NET_POLL_FAILED;
|
|
|
|
}
|
2021-07-19 21:55:20 +00:00
|
|
|
ret = 0;
|
|
|
|
if (FD_ISSET(fd, &read_fds)) ret |= MBEDTLS_NET_POLL_READ;
|
|
|
|
if (FD_ISSET(fd, &write_fds)) ret |= MBEDTLS_NET_POLL_WRITE;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Portable usleep helper
|
|
|
|
*
|
|
|
|
* \param usec Amount of microseconds to sleep
|
|
|
|
*
|
|
|
|
* \note Real amount of time slept will not be less than
|
|
|
|
* select()'s timeout granularity (typically, 10ms).
|
|
|
|
*/
|
|
|
|
void mbedtls_net_usleep(unsigned long usec) {
|
|
|
|
usleep(usec);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Read at most 'len' characters. If no error occurs,
|
|
|
|
* the actual amount read is returned.
|
|
|
|
*
|
|
|
|
* \param ctx Socket
|
|
|
|
* \param buf The buffer to write to
|
|
|
|
* \param len Maximum length of the buffer
|
|
|
|
*
|
|
|
|
* \return the number of bytes received,
|
|
|
|
* or a non-zero error code; with a non-blocking socket,
|
|
|
|
* MBEDTLS_ERR_SSL_WANT_READ indicates read() would block.
|
|
|
|
*/
|
|
|
|
int mbedtls_net_recv(void *ctx, unsigned char *buf, size_t len) {
|
|
|
|
int ret = MBEDTLS_ERR_THIS_CORRUPTION;
|
|
|
|
int fd = ((mbedtls_net_context *)ctx)->fd;
|
|
|
|
if (fd < 0) return MBEDTLS_ERR_NET_INVALID_CONTEXT;
|
|
|
|
ret = (int)read(fd, buf, len);
|
|
|
|
if (ret < 0) {
|
|
|
|
if (net_would_block(ctx) != 0) return MBEDTLS_ERR_SSL_WANT_READ;
|
|
|
|
if (errno == EPIPE || errno == ECONNRESET)
|
|
|
|
return MBEDTLS_ERR_NET_CONN_RESET;
|
|
|
|
if (errno == EINTR) return MBEDTLS_ERR_SSL_WANT_READ;
|
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 (errno == ECANCELED) return MBEDTLS_ERR_SSL_CANCELED;
|
2021-07-19 21:55:20 +00:00
|
|
|
return MBEDTLS_ERR_NET_RECV_FAILED;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Read at most 'len' characters, blocking for at most
|
|
|
|
* 'timeout' seconds. If no error occurs, the actual amount
|
|
|
|
* read is returned.
|
|
|
|
*
|
|
|
|
* \note The current implementation of this function uses
|
|
|
|
* select() and returns an error if the file descriptor
|
|
|
|
* is \c FD_SETSIZE or greater.
|
|
|
|
*
|
|
|
|
* \param ctx Socket
|
|
|
|
* \param buf The buffer to write to
|
|
|
|
* \param len Maximum length of the buffer
|
|
|
|
* \param timeout Maximum number of milliseconds to wait for data
|
|
|
|
* 0 means no timeout (wait forever)
|
|
|
|
*
|
|
|
|
* \return The number of bytes received if successful.
|
|
|
|
* MBEDTLS_ERR_SSL_TIMEOUT if the operation timed out.
|
|
|
|
* MBEDTLS_ERR_SSL_WANT_READ if interrupted by a signal.
|
|
|
|
* Another negative error code (MBEDTLS_ERR_NET_xxx)
|
|
|
|
* for other failures.
|
|
|
|
*
|
|
|
|
* \note This function will block (until data becomes available or
|
|
|
|
* timeout is reached) even if the socket is set to
|
|
|
|
* non-blocking. Handling timeouts with non-blocking reads
|
|
|
|
* requires a different strategy.
|
|
|
|
*/
|
|
|
|
int mbedtls_net_recv_timeout(void *ctx, unsigned char *buf, size_t len,
|
|
|
|
uint32_t timeout) {
|
|
|
|
int ret = MBEDTLS_ERR_THIS_CORRUPTION;
|
|
|
|
struct timeval tv;
|
|
|
|
fd_set read_fds;
|
|
|
|
int fd = ((mbedtls_net_context *)ctx)->fd;
|
|
|
|
if (fd < 0) return MBEDTLS_ERR_NET_INVALID_CONTEXT;
|
|
|
|
/* A limitation of select() is that it only works with file descriptors
|
|
|
|
* that are strictly less than FD_SETSIZE. This is a limitation of the
|
|
|
|
* fd_set type. Error out early, because attempting to call FD_SET on a
|
|
|
|
* large file descriptor is a buffer overflow on typical platforms. */
|
|
|
|
if (fd >= FD_SETSIZE) return (MBEDTLS_ERR_NET_POLL_FAILED);
|
|
|
|
FD_ZERO(&read_fds);
|
|
|
|
FD_SET(fd, &read_fds);
|
|
|
|
tv.tv_sec = timeout / 1000;
|
|
|
|
tv.tv_usec = (timeout % 1000) * 1000;
|
|
|
|
ret = select(fd + 1, &read_fds, NULL, NULL, timeout == 0 ? NULL : &tv);
|
|
|
|
/* Zero fds ready means we timed out */
|
|
|
|
if (ret == 0) return MBEDTLS_ERR_SSL_TIMEOUT;
|
|
|
|
if (ret < 0) {
|
|
|
|
if (errno == EINTR) return MBEDTLS_ERR_SSL_WANT_READ;
|
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 (errno == ECANCELED) return MBEDTLS_ERR_SSL_CANCELED;
|
2021-07-19 21:55:20 +00:00
|
|
|
return MBEDTLS_ERR_NET_RECV_FAILED;
|
|
|
|
}
|
|
|
|
/* This call will not block */
|
|
|
|
return mbedtls_net_recv(ctx, buf, len);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Write at most 'len' characters. If no error occurs,
|
|
|
|
* the actual amount read is returned.
|
|
|
|
*
|
|
|
|
* \param ctx Socket
|
|
|
|
* \param buf The buffer to read from
|
|
|
|
* \param len The length of the buffer
|
|
|
|
*
|
|
|
|
* \return the number of bytes sent,
|
|
|
|
* or a non-zero error code; with a non-blocking socket,
|
|
|
|
* MBEDTLS_ERR_SSL_WANT_WRITE indicates write() would block.
|
|
|
|
*/
|
|
|
|
int mbedtls_net_send(void *ctx, const unsigned char *buf, size_t len) {
|
|
|
|
int ret = MBEDTLS_ERR_THIS_CORRUPTION;
|
|
|
|
int fd = ((mbedtls_net_context *)ctx)->fd;
|
|
|
|
if (fd < 0) return MBEDTLS_ERR_NET_INVALID_CONTEXT;
|
|
|
|
ret = (int)write(fd, buf, len);
|
|
|
|
if (ret < 0) {
|
|
|
|
if (net_would_block(ctx) != 0) return MBEDTLS_ERR_SSL_WANT_WRITE;
|
|
|
|
if (errno == EPIPE || errno == ECONNRESET)
|
|
|
|
return MBEDTLS_ERR_NET_CONN_RESET;
|
|
|
|
if (errno == EINTR) return MBEDTLS_ERR_SSL_WANT_WRITE;
|
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 (errno == ECANCELED) return MBEDTLS_ERR_SSL_CANCELED;
|
2021-07-19 21:55:20 +00:00
|
|
|
return MBEDTLS_ERR_NET_SEND_FAILED;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Closes down the connection and free associated data
|
|
|
|
*
|
|
|
|
* \param ctx The context to close
|
|
|
|
*/
|
|
|
|
void mbedtls_net_close(mbedtls_net_context *ctx) {
|
|
|
|
if (ctx->fd == -1) return;
|
|
|
|
close(ctx->fd);
|
|
|
|
ctx->fd = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* \brief Gracefully shutdown the connection and free associated data
|
|
|
|
*
|
|
|
|
* \param ctx The context to free
|
|
|
|
*/
|
|
|
|
void mbedtls_net_free(mbedtls_net_context *ctx) {
|
|
|
|
if (ctx->fd == -1) return;
|
|
|
|
shutdown(ctx->fd, 2);
|
|
|
|
close(ctx->fd);
|
|
|
|
ctx->fd = -1;
|
|
|
|
}
|