PhysFrames mapped multiple times?

[pitust]

Yeah, stuff gets mapped multiple times. Here is my log output [from an isa-debugcon]:

; STUFF HERE
MAP @ PhysFrame[4KiB](0x5f8000) to Page[4KiB](0xf001f7000)
Now mapping Seg
MAP @ PhysFrame[4KiB](0x5f8000) to Page[4KiB](0xf001f7000)
panicked at 'kernel mapping failed: Mapping(PageAlreadyMapped(PhysFrame[4KiB](0x5f8000)))', src/main.rs:268:6

And it just hangs. So... It seems that sections do NOT have to be 4k-aligned in ELFs. The culprit in my case was .got. Forcing it to stay aligned fixed my problem. But, this should just work and not require custom linker scripts.

[pitust]

I have some extra logging output from the bootloader

[phil-opp]

Are you using any linker arguments? For me, the default linker script without any arguments seems to always produce aligned segments.

Either way, I agree that it would be nice if the bootloader could work with non-aligned segments.

[pitust]

Indeed, my target triple is:

{
    "llvm-target": "x86_64-unknown-none",
    "data-layout": "e-m:e-i64:64-f80:128-n8:16:32:64-S128",
    "arch": "x86_64",
    "target-endian": "little",
    "target-pointer-width": "64",
    "target-c-int-width": "32",
    "os": "none",
    "executables": true,
    "linker-flavor": "ld.lld",
    "linker": "rust-lld",
    "disable-redzone": true,
    "features": "-mmx,+sse",
    "panic-strategy": "unwind",
    "pre-link-args": {
        "ld.lld": [
            "/opt/cross/lib/gcc/x86_64-elf/10.2.0/libgcc.a", "--allow-multiple-definition",
        ]
    }
}

I need libgcc for _Unwind_* functions (i use that for stack traces). I assume that it is due to libgcc.a having a .got section which isn't 4k-aligned.

[pitust]

anyway i switched to multiboot2 and first edition of the guide, because i want support for modern architectures.

[pitust]

Wait, no, i got wrong args, i do indeed have a linker script, the one found on osdev.org (https://wiki.osdev.org/Bare_Bones)

[phil-opp]

i do indeed have a linker script, the one found on osdev.org (https://wiki.osdev.org/Bare_Bones)

Ah, that explains things. I think most default linker scripts align all sections properly, but a custom linker script might not. The linker script you linked does not mention any .got section, so the linker just places it after the last section, without any special alignment.

While I still agree that it would be nice if the bootloader crate supported non-aligned sections, section alignment is still recommended because otherwise you might not be able to set the correct write/executable permissions for these pages. For example, if .text shares a page with .data, that page must be mapped as both writable and executable, which is generally discouraged since attackers can easily run their injected code this way.

anyway i switched to multiboot2 and first edition of the guide, because i want support for modern architectures.

Could you clarify what you mean with "modern architectures"? Support for UEFI booting or something else? I'm currently working on a new version of the bootloader with UEFI support, so it would be useful to know if something else is missing too.

[pitust]

UEFI and coreboot. I know CSM is a thing, but i wouldn't rely on that. (there are some laptops that use coreboot out of the box, most notably pixelbooks and everything with System76 Open Firmware)