I meant with PROT_EXEC flag.

Yep and with PROT_READ

I am working on project similar case with you. Our JIT-ed code needs to be something like that.

Yep and with PROT_READ

PROT_WRITE is more relevant though.

PROT_WRITE is more relevant though.

Do I need that?

Yes, to load the code to mapped memory.

Oh.

Yes, to load the code to mapped memory.

I load the mapped memory from a file

Depends, if you open() and put the fd to mmap(), then no need PROT_WRITE.

I load the mapped memory from a file

Ah right.

👌

Btw @tuxifan, what IDE do you use? I am looking for __AARGS__ on google, it always brings me to __VA_ARGS__ instead.

Btw @tuxifan, what IDE do you use? I am looking for __AARGS__ on google, it always brings me to __VA_ARGS__ instead.

qt creator

Hmm, still can't find it with "qt creator" keyword.

__AARGS__ does not seem to be a built in macro in gcc 9.3.0 Linux x86_64.

Dear admin, can I appeal to remove this warning? Is there really no tolerance to post a message link?

Dear admin, can I appeal to remove this warning? Is there really no tolerance to post a message link?

What did you post?

ssize_t my_write(int edi, const void *rsi, size_t rdx) { ssize_t ret; __asm__ volatile( "syscall" : "=a"(ret) : "D"(edi), "S"(rsi), "d"(rdx), "a"(1) : "memory" ); return ret; }

I posted a link to this message.

I posted a link to this message.

Lol

Removed

Thanks :D

how do I create a new thread without libc on linux?

how do I create a new thread without libc on linux?

man 2 clone

thx

how do I create a new thread without libc on linux?

clone syscall

clone syscall

The stack argument specifies the location of the stack used by the child process. Since the child and calling process may share memory, it is not possible for the child process to execute in the same stack as the calling process. But how do I allocate that stack without malloc()?

The stack argument specifies the location of the stack used by the child process. Since the child and calling process may share memory, it is not possible for the child process to execute in the same stack as the calling process. But how do I allocate that stack without malloc()?

Using sbrk or mmap (the way malloc works)

The stack argument specifies the location of the stack used by the child process. Since the child and calling process may share memory, it is not possible for the child process to execute in the same stack as the calling process. But how do I allocate that stack without malloc()?

You should not allocate the stack with malloc.

Stack is VM_GROWSDOWN.

It should be allocated by using mmap.

Using sbrk or mmap (the way malloc works)

sbrk() is not available either… what do I need to pass to mmap?

Any special stuff?

Wait a minute.

sbrk() is not available either… what do I need to pass to mmap?

mmap(NULL, 8392704, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_STACK, -1, 0)

That is the call to mmap by pthread.

mmap(NULL, 8392704, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_STACK, -1, 0)

thx

clone(payload_main, newstack, CLONE_FILES | CLONE_FS | CLONE_IO | CLONE_PTRACE | CLONE_VM, NULL); Like that?

clone(payload_main, newstack, CLONE_FILES | CLONE_FS | CLONE_IO | CLONE_PTRACE | CLONE_VM, NULL); Like that?

Not really sure, but the clone arguments are different.

Please try to inspect the pthread call.

#include <pthread.h> void *test(void *x) { return NULL; } int main() { pthread_t th; pthread_create(&th, NULL, test, NULL); pthread_join(th, NULL); }

Not really sure, but the clone arguments are different.

clone(int (*fn)(void *), void *stack, int flags, void *arg)

#include <pthread.h> void *test(void *x) { return NULL; } int main() { pthread_t th; pthread_create(&th, NULL, test, NULL); pthread_join(th, NULL); }

what does pthread have to do with clone?

what does pthread have to do with clone?

It uses clone to spawn the thread.

You can follow how it creates the thread.

clone(payload_main, newstack, CLONE_FILES | CLONE_FS | CLONE_IO | CLONE_PTRACE | CLONE_VM, NULL); Like that?

clone(child_stack=0x7fff52447380, flags=CLONE_VM|CLONE_PIDFD|CLONE_PTRACE|CLONE_VFORK|CLONE_NEWNS|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID|CLONE_NEWCGROUP|CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWPID|0x7f4200400000|71, parent_tid=0x80002700, tls=0x3, child_tidptr=0x2) = -1 EINVAL (Invalid argument)

clone(int (*fn)(void *), void *stack, int flags, void *arg)

but man says this

I have never worked with raw thread like that. Don't really know, but investigating the pthread and follow it may be the better way.

I have never worked with raw thread like that. Don't really know, but investigating the pthread and follow it may be the better way.

I am not able to use pthread here

I am not able to use pthread here

I meant, inspect the syscall and arguments done by pthread.

It is using the syscall internally too.

I meant, inspect the syscall and arguments done by pthread.

but where to pass the childs entry function?

I am still looking for that too.

but where to pass the childs entry function?

Yeah, it is the first argument.

Yeah, it is the first argument.

but the first argument is child_stack??

but the first argument is child_stack??

It seems clone by #include <sched.h> is different with literal clone syscall.

It seems clone by #include <sched.h> is different with literal clone syscall.

ouch

ouch

The important part: The raw clone() system call corresponds more closely to fork(2) in that execution in the child continues from the point of the call.

I have no idea what TLS is

or TID

I have no idea what TLS is

The TLS (Thread Local Storage) descriptor is set to tls.

It seems similar to Thread Safe Resource Manager stuff.

Quoted from man: The TLS (Thread Local Storage) descriptor is set to tls. The interpretation of tls and the resulting effect is architecture dependent. On x86, tls is interpreted as a struct user_desc * (see set_thread_area(2)). On x86-64 it is the new value to be set for the %fs base register (see the ARCH_SET_FS argument to arch_prctl(2)). On architectures with a dedicated TLS register, it is the new value of that register. Use of this flag requires detailed knowledge and generally it should not be used except in libraries implementing threading.

The TLS (Thread Local Storage) descriptor is set to tls.

what if I need my thread to be in the same stack?

what if I need my thread to be in the same stack?

Your thread will overwrite your main stack, this clearly will mess your program.

Your thread will overwrite your main stack, this clearly will mess your program.

Hmm, do I really have to care about that?

Yes, every calls to a function store the return address into the stack.

Your thread will call functions. So it is a problem if your thread uses main stack.

which is bad Ig?

which is bad Ig?

Of course very bad.

Your return statement won't work correctly.

Your local variables too.

The TLS (Thread Local Storage) descriptor is set to tls.

And what do I have to set it to?

Of course very bad.

clone(child_stack=0x7ffdf36a0110, flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_PTRACE|CLONE_IO) = 67238 strace: Detached unknown pid 67238 write(1, "Failed\n", 7 <unfinished ...> +++ killed by SIGSEGV +++ I see...

And what do I have to set it to?

I can't find the simple example of this.

I can't find the simple example of this.

:-/

I need to learn more about this, it is very complicated to me to make a conclusion from pthread sources.

So this is where I am stuck now.

So this is where I am stuck now.

char newstack[1024]; will get removed when return.