/* $Id: bs3kit-docs.c 60600 2016-04-20 13:16:20Z vboxsync $ */ /** @file * BS3Kit - Documentation. */ /* * Copyright (C) 2007-2016 Oracle Corporation * * This file is part of VirtualBox Open Source Edition (OSE), as * available from http://www.virtualbox.org. This file is free software; * you can redistribute it and/or modify it under the terms of the GNU * General Public License (GPL) as published by the Free Software * Foundation, in version 2 as it comes in the "COPYING" file of the * VirtualBox OSE distribution. VirtualBox OSE is distributed in the * hope that it will be useful, but WITHOUT ANY WARRANTY of any kind. * * The contents of this file may alternatively be used under the terms * of the Common Development and Distribution License Version 1.0 * (CDDL) only, as it comes in the "COPYING.CDDL" file of the * VirtualBox OSE distribution, in which case the provisions of the * CDDL are applicable instead of those of the GPL. * * You may elect to license modified versions of this file under the * terms and conditions of either the GPL or the CDDL or both. */ /** @page pg_bs3kit BS3Kit - Boot Sector Kit \#3 * * The BS3Kit is a framework for bare metal floppy/usb image tests. * * The 3rd iteration of the framework includes support for 16-bit and 32-bit * C/C++ code, with provisions for 64-bit C code to possibly be added later. * The C code have to do without a runtime library, otherwhat what we can share * possibly with IPRT. * * This iteration also adds a real linker into the picture, which is an * improvment over early when all had to done in a single assembler run with * lots of includes and macros controlling what we needed. The functions are no * in separate files and compiled/assembled into libraries, so the linker will * only include exactly what is needed. The current linker is the OpenWatcom * one, wlink, that we're already using when building the BIOSes. If it wasn't * for the segment/selector fixups in 16-bit code (mostly), maybe we could * convince the ELF linker from GNU binutils to do the job too (with help from * the ). * * * @sa grp_bs3kit, grp_bs3kit_tmpl, grp_bs3kit_cmn, grp_bs3kit_mode, * grp_bs3kit_system * * @section sec_calling_convention Calling convention * * Because we're not mixing with C code, we will use __cdecl for 16-bit and * 32-bit code, where as 64-bit code will use the microsoft calling AMD64 * convention. To avoid unnecessary %ifdef'ing in assembly code, we will use a * macro to load the RCX, RDX, R8 and R9 registers off the stack in 64-bit * assembly code. * * Register treatment in 16-bit __cdecl, 32-bit __cdecl and 64-bit msabi: * * | Register | 16-bit | 32-bit | 64-bit | ASM template | * | ------------ | ----------- | ---------- | --------------- | ------------ | * | EAX, RAX | volatile | volatile | volatile | volatile | * | EBX, RBX | volatile | preserved | preserved | both | * | ECX, RCX | volatile | volatile | volatile, arg 0 | volatile | * | EDX, RDX | volatile | volatile | volatile, arg 1 | volatile | * | ESP, RSP | preserved | preserved | preserved | preserved | * | EBP, RBP | preserved | preserved | preserved | preserved | * | EDI, RDI | preserved | preserved | preserved | preserved | * | ESI, RSI | preserved | preserved | preserved | preserved | * | R8 | volatile | volatile | volatile, arg 2 | volatile | * | R9 | volatile | volatile | volatile, arg 3 | volatile | * | R10 | volatile | volatile | volatile | volatile | * | R11 | volatile | volatile | volatile | volatile | * | R12 | volatile | volatile | preserved | preserved(*) | * | R13 | volatile | volatile | preserved | preserved(*) | * | R14 | volatile | volatile | preserved | preserved(*) | * | R15 | volatile | volatile | preserved | preserved(*) | * | RFLAGS.DF | =0 | =0 | =0 | =0 | * | CS | preserved | preserved | preserved | preserved | * | DS | preserved! | preserved? | preserved | both | * | ES | volatile | volatile | preserved | volatile | * | FS | preserved | preserved | preserved | preserved | * | GS | preserved | volatile | preserved | both | * | SS | preserved | preserved | preserved | preserved | * * The 'both' here means that we preserve it wrt to our caller, while at the * same time assuming anything we call will clobber it. * * The 'preserved(*)' marking of R12-R15 indicates that they'll be preserved in * 64-bit mode, but may be changed in certain cases when running 32-bit or * 16-bit code. This is especially true if switching CPU mode, e.g. from 32-bit * protected mode to 32-bit long mode. * * Return values are returned in the xAX register, but with the following * caveats for values larger than ARCH_BITS: * - 16-bit code: * - 32-bit values are returned in AX:DX, where AX holds bits 15:0 and * DX bits 31:16. * - 64-bit values are returned in DX:CX:BX:AX, where DX holds bits * 15:0, CX bits 31:16, BX bits 47:32, and AX bits 63:48. * - 32-bit code: * - 64-bit values are returned in EAX:EDX, where eax holds the least * significant bits. * * The DS segment register is pegged to BS3DATA16_GROUP in 16-bit code so that * we don't need to reload it all the time. This allows us to modify it in * ring-0 and mode switching code without ending up in any serious RPL or DPL * trouble. In 32-bit and 64-bit mode the DS register is a flat, unlimited, * writable selector. * * In 16-bit and 32-bit code we do not assume anything about ES, FS, and GS. * * * For an in depth coverage of x86 and AMD64 calling convensions, see * http://homepage.ntlworld.com/jonathan.deboynepollard/FGA/function-calling-conventions.html * * * * @section sec_modes Execution Modes * * BS3Kit defines a number of execution modes in order to be able to test the * full CPU capabilities (that VirtualBox care about anyways). It currently * omits system management mode, hardware virtualization modes, and security * modes as those aren't supported by VirtualBox or are difficult to handle. * * The modes are categorized into normal and weird ones. * * The normal ones are: * + RM - Real mode. * + PE16 - Protected mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PE32 - Protected mode running 32-bit code, 32-bit TSS and 32-bit handlers. * + PEV86 - Protected mode running v8086 code, 32-bit TSS and 32-bit handlers. * + PP16 - 386 paged mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PP32 - 386 paged mode running 32-bit code, 32-bit TSS and 32-bit handlers. * + PPV86 - 386 paged mode running v8086 code, 32-bit TSS and 32-bit handlers. * + PAE16 - PAE paged mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PAE32 - PAE paged mode running 32-bit code, 32-bit TSS and 32-bit handlers. * + PAEV86 - PAE paged mode running v8086 code, 32-bit TSS and 32-bit handlers. * + LM16 - AMD64 long mode running 16-bit code, 64-bit TSS and 64-bit handlers. * + LM32 - AMD64 long mode running 32-bit code, 64-bit TSS and 64-bit handlers. * + LM64 - AMD64 long mode running 64-bit code, 64-bit TSS and 64-bit handlers. * * The weird ones: * + PE16_32 - Protected mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PE16_V86 - Protected mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PE32_16 - Protected mode running 32-bit code, 32-bit TSS and 32-bit handlers. * + PP16_32 - 386 paged mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PP16_V86 - 386 paged mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PP32_16 - 386 paged mode running 32-bit code, 32-bit TSS and 32-bit handlers. * + PAE16_32 - PAE paged mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PAE16_V86 - PAE paged mode running 16-bit code, 16-bit TSS and 16-bit handlers. * + PAE32_16 - PAE paged mode running 32-bit code, 32-bit TSS and 32-bit handlers. * * Actually, the PE32_16, PP32_16 and PAE32_16 modes aren't all that weird and fits in * right next to LM16 and LM32, but this is the way it ended up. :-) * */