Reachable unwrap panic in sys_munmap() at legacy.rs

[Marsman1996]

Describe the bug

There is a reachable unwrap panic in sys_munmap() caused by aligning large len (i.e., 0xffffffffffffffff) of a munmap syscall.

ruxos/api/ruxos_posix_api/src/imp/mmap/legacy.rs

Line 72 in b1f880b

let layout = Layout::from_size_align(len, 8).unwrap();

To Reproduce

1. Compile a program which calls munmap:

#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <unistd.h>

int main()
{
    void *addr = mmap(NULL, 4096, 0x3, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
    munmap(addr, 0xffffffffffffffff);

    return 0;
}

With following features:

alloc 
irq
musl
multitask
fs
pipe
poll
rtc
signal
virtio-9p

2. Run the compiled program in RuxOS.

Expected behavior

RuxOS reports panic and is terminated.

Environment

• RuxOS version: main b1f880b
• ubuntu:22.04 in Docker
• 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz

Logs

[1728388549.205894 0:2 ruxos_posix_api::imp::mmap::legacy:67] sys_munmap <= start: 0xffffff80003fe000, len: 18446744073709551615
[1728388549.206739 0:2 ruxruntime::lang_items:14] panicked at api/ruxos_posix_api/src/imp/mmap/legacy.rs:72:54:
called `Result::unwrap()` on an `Err` value: LayoutError
[1728388549.207796 0:2 ruxhal::platform::x86_pc::misc:16] Shutting down...

[Marsman1996]

Currently RuxOS will panic when page fault happening. For example, program access an invalid memory region.

I'm wondering is it a feature, or will RuxOS handle it in the future?

[coolyjg]

Thanks for your bug report for RuxOS!
It is not a feature or anything. We directly used unwrap() for Layout::from_size_align and did not handle Err value, and finally caused this issue. @ken4647
By the way, there are some suggestions for this issue:

1. munmap should fail and return error number when some part of the region being unmapped is not part of the currently valid address space(see system call manual).
2. sys_munmap panics due to large len, which should be checked.

[Marsman1996]

Thank you for your quick reply!

And my question is actually about this:

ruxos/modules/ruxhal/src/arch/x86_64/trap.rs

Lines 54 to 57 in b1f880b

	panic!(
	"Kernel #PF @ {:#x}, fault_vaddr={:#x}, error_code={:#x}:\n{:#x?}",
	tf.rip, vaddr, tf.error_code, tf,
	);

When RuxOS accesses an invalid memory and causes a page fault, it will panic.

Will it be handled in the future?
Since it is an unikenel and run one program at once, it seems ok to panic directly?

[coolyjg]

When accessing invalid memory, program behavior is undefined, so panic is a safe implementation. (I'm not sure whether it will be handled in the future.)

[Marsman1996]

And a similarly unwrap panic for sys_mmap() at legacy.rs.

ruxos/api/ruxos_posix_api/src/imp/mmap/legacy.rs

Line 44 in b1f880b

let layout = Layout::from_size_align(len, 8).unwrap();

If we run following program which calls mmap with large size (i.e., 0x7ffffffffffffffe)

#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <unistd.h>

int main()
{
    mmap(NULL, 0x7ffffffffffffffe, 0x3, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);

    return 0;
}

With following features:

alloc 
irq
musl
multitask
pipe
poll
rtc
signal
virtio-9p

logs like

[1728436128.205115 0:2 ruxmusl::x86_64:10] x86 syscall <= syscall_name: MMAP
[1728436128.205642 0:2 ruxos_posix_api::imp::mmap::legacy:32] sys_mmap <= start: 0x0, len: 9223372036854775806, fd: -1
[1728436128.206396 0:2 ruxruntime::lang_items:14] panicked at api/ruxos_posix_api/src/imp/mmap/legacy.rs:44:54:
called `Result::unwrap()` on an `Err` value: LayoutError
[1728436128.207400 0:2 ruxhal::platform::x86_pc::misc:16] Shutting down...