1
Etherboot/NILO i386 initialisation path and external call interface
2
===================================================================
6
GCC compiles 32-bit code. It is capable of producing
7
position-independent code, but the resulting binary is about 25%
8
bigger than the corresponding fixed-position code. Since one main use
9
of Etherboot is as firmware to be burned into an EPROM, code size must
10
be kept as small as possible.
12
This means that we want to compile fixed-position code with GCC, and
13
link it to have a predetermined start address. The problem then is
14
that we must know the address that the code will be loaded to when it
15
runs. There are several ways to solve this:
17
1. Pick an address, link the code with this start address, then make
18
sure that the code gets loaded at that location. This is
19
problematic, because we may pick an address that we later end up
20
wanting to use to load the operating system that we're booting.
22
2. Pick an address, link the code with this start address, then set up
23
virtual addressing so that the virtual addresses match the
24
link-time addresses regardless of the real physical address that
25
the code is loaded to. This enables us to relocate Etherboot to
26
the top of high memory, where it will be out of the way of any
27
loading operating system.
29
3. Link the code with a text start address of zero and a data start
30
address also of zero. Use 16-bit real mode and the
31
quasi-position-independence it gives you via segment addressing.
32
Doing this requires that we generate 16-bit code, rather than
33
32-bit code, and restricts us to a maximum of 64kB in each segment.
35
There are other possible approaches (e.g. including a relocation table
36
and code that performs standard dynamic relocation), but the three
37
options listed above are probably the best available.
39
Etherboot can be invoked in a variety of ways (ROM, floppy, as a PXE
40
NBP, etc). Several of these ways involve control being passed to
41
Etherboot with the CPU in 16-bit real mode. Some will involve the CPU
42
being in 32-bit protected mode, and there's an outside chance that
43
some may involve the CPU being in 16-bit protected mode. We will
44
almost certainly have to effect a CPU mode change in order to reach
45
the mode we want to be in to execute the C code.
47
Additionally, Etherboot may wish to call external routines, such as
48
BIOS interrupts, which must be called in 16-bit real mode. When
49
providing a PXE API, Etherboot must provide a mechanism for external
50
code to call it from 16-bit real mode.
52
Not all i386 builds of Etherboot will want to make real-mode calls.
53
For example, when built for LinuxBIOS rather than the standard PCBIOS,
54
no real-mode calls are necessary.
56
For the ultimate in PXE compatibility, we may want to build Etherboot
57
to run permanently in real mode.
59
There is a wide variety of potential combinations of mode switches
60
that we may wish to implement. There are additional complications,
61
such as the inability to access a high-memory stack when running in
64
2. Transition libraries
66
To handle all these various combinations of mode switches, we have
67
several "transition" libraries in Etherboot. We also have the concept
68
of an "internal" and an "external" environment. The internal
69
environment is the environment within which we can execute C code.
70
The external environment is the environment of whatever external code
71
we're trying to interface to, such as the system BIOS or a PXE NBP.
73
As well as having a separate addressing scheme, the internal
74
environment also has a separate stack.
76
The transition libraries are:
80
librm handles transitions between an external 16-bit real-mode
81
environment and an internal 32-bit protected-mode environment with
86
libkir handles transitions between an external 16-bit real-mode (or
87
16:16 or 16:32 protected-mode) environment and an internal 16-bit
88
real-mode (or 16:16 protected-mode) environment.
92
libpm handles transitions between an external 32-bit protected-mode
93
environment with flat physical addresses and an internal 32-bit
94
protected-mode environment with virtual addresses.
96
The transition libraries handle the transitions required when
97
Etherboot is started up for the first time, the transitions required
98
to execute any external code, and the transitions required when
99
Etherboot exits (if it exits). When Etherboot provides a PXE API,
100
they also handle the transitions required when a PXE client makes a
101
PXE API call to Etherboot.
103
Etherboot may use multiple transition libraries. For example, an
104
Etherboot ELF image does not require librm for its initial transitions
105
from prefix to runtime, but may require librm for calling external
108
3. Setup and initialisation
110
Etherboot is conceptually divided into the prefix, the decompressor,
111
and the runtime image. (For non-compressed images, the decompressor
112
is a no-op.) The complete image comprises all three parts and is
113
distinct from the runtime image, which exclude the prefix and the
116
The prefix does several tasks:
118
Load the complete image into memory. (For example, the floppy
119
prefix issues BIOS calls to load the remainder of the complete image
120
from the floppy disk into RAM, and the ISA ROM prefix copies the ROM
121
contents into RAM for faster access.)
123
Call the decompressor, if the runtime image is compressed. This
124
decompresses the runtime image.
126
Call the runtime image's setup() routine. This is a routine
127
implemented in assembly code which sets up the internal environment
128
so that C code can execute.
130
Call the runtime image's arch_initialise() routine. This is a
131
routine implemented in C which does some basic startup tasks, such
132
as initialising the console device, obtaining a memory map and
133
relocating the runtime image to high memory.
135
Call the runtime image's arch_main() routine. This records the exit
136
mechanism requested by the prefix and calls main(). (The prefix
137
needs to register an exit mechanism because by the time main()
138
returns, the memory occupied by the prefix has most likely been
141
When acting as a PXE ROM, the ROM prefix contains an UNDI loader
142
routine in addition to its usual code. The UNDI loader performs a
143
similar sequence of steps:
145
Load the complete image into memory.
147
Call the decompressor.
149
Call the runtime image's setup() routine.
151
Call the runtime image's arch_initialise() routine.
153
Call the runtime image's install_pxe_stack() routine.
157
The runtime image's setup() routine will perform the following steps:
159
Switch to the internal environment using an appropriate transition
160
library. This will record the parameters of the external
163
Set up the internal environment: load a stack, and set up a GDT for
164
virtual addressing if virtual addressing is to be used.
166
Switch back to the external environment using the transition
167
library. This will record the parameters of the internal
170
Once the setup() routine has returned, the internal environment has been
171
set up ready for C code to run. The prefix can call C routines using
172
a function from the transition library.
174
The runtime image's arch_initialise() routine will perform the
179
Initialise the console device(s) and print a welcome message.
181
Obtain a memory map via the INT 15,E820 BIOS call or suitable
182
fallback mechanism. [not done if libkir is being used]
184
Relocate the runtime image to the top of high memory. [not done if
185
libkir is being used]
187
Install librm to base memory. [done only if librm is being used]
191
Return to the prefix, setting registers to indicate to the prefix
192
the new location of the transition library, if applicable. Which
193
registers these are is specific to the transition library being
196
Once the arch_initialise() routine has returned, the prefix will
197
probably call arch_main().