Re: Etherboot 5.5

Michael Brown wrote:

> New features that are already in place include:
> 
>   An improved build system:
> 
>     Changing configuration options no longer requires rebuilding the 
>     entire Etherboot code tree.
> 
>     All code is compiled.  Only the objects we want are linked in.  This 
>     reduces #ifdef spaghetti while keeping binary sizes small.  See 
>     http://www.etherboot.org/api/ifdef_harmful.html for details.

Very good, the sheer number of configurations was mind boggling.

>     Standardised debugging.  To enable debugging of a particular component 
>     (e.g. the PCI bus), simply append DEBUG=pci to the command line, e.g.
>     "make bin/rtl8139.dsk DEBUG=pci"

This is very nice, it was one of the things I have been thinking of for 
some time.  I was thinking of it at the nic driver level but you have 
considerable expanded the scope.

>     Arbitrary alignment (via __attribute__((aligned(x))) is supported and 
>     will be honoured by the relocation code.

Cool, I see you fixed the r8169...

>     Adding new features (e.g. new drivers, new download protocols, new 
>     console types) requires *no* changes to the core code, or to files 
>     such as "Families".  Just drop in the .c file and it will work.

Very nice, addind drivers and protocols was aleardy easy, this is 
incredible. ;-)

>   The all-drivers image has been reduced in size by about 52%, simply by 
>   sharing buffer space between drivers.

Great!

>   Support for a much wider range of buses.  We now support PCI, ISAPnP, 
>   EISA, MCA and ISA as individual buses, rather than pretending that 
>   everything is either PCI or plain old ISA.  ISAPnP code works for cards 
>   other than the 3c515.

Very good.  The ISAPnP was always a hack.  I was suprised when I learned 
it worked for other people as it was all about me and my 486 when I put 
it together.

>   A concept of "errno", allowing for proper PXE status codes and more 
>   user-friendly error messages.
> 
>   A new real-mode call interface (again!).
> 
>   The ability to operate completely in 16-bit real mode, never switching 
>   to protected mode.  This is necessary in order to comply fully with the 
>   PXE "specification".  This is controlled by a compile-time option 
>   -DKEEP_IT_REAL; off by default.

I understand that this is one of the more difficult parts of the PXE 
spec.  Doesn't it contradict some other requirement in the spec (not 
that that s unusual?

>   Automatically-generated internal API documentation via doxygen.  See
>   http://www.etherboot.org/api/files.html

I browsed this a little today.  I am looking at using doxygen at work so 
this is timely.  I must brush up on what needs to be done and update 
some of the driver documentation.

Wow!!  You have been busy.  I browsed the cvs via the web and the sheer 
number of changes are amazing.  You must not have slept since Christmas.

> Of course, none of this comes for free.  The following features are 
> currently broken:

Not to worry, I am sure most of it will get fixed fairly quickly.

>   All architectures except for i386.  If anyone wants to donate ia64, 
>   armnommu, or e1 hardware to me, then I can do something about this 
>   myself.

I don't know if e1 even worked in 5.2.

> To forestall the inevitable questions: yes, it was worth breaking all 
> this.  The internal architecture of Etherboot had become far too 
> convoluted, and was in dire need of being torn apart and reassembled.

Such a huge amount of work does not come without issues.  Each of the 
last number of versions of Etherboot have broken a lot of things.  They 
get fixed...

BTW, do we need an updated compiler/build tools?  I received the 
following message from the first build I tried:

In file included from arch/i386/core/hooks.c:1:
arch/i386/include/registers.h:19: warning: declaration does not declare 
anything
arch/i386/include/registers.h:87: warning: declaration does not declare 
anything
arch/i386/include/registers.h:88: warning: declaration does not declare 
anything
In file included from include/etherboot.h:157,
                  from include/buffer.h:4,
                  from include/dev.h:6,
                  from include/main.h:4,
                  from arch/i386/core/hooks.c:2:
include/tftp.h:88: warning: "struct buffer" declared inside parameter list
include/tftp.h:88: warning: its scope is only this definition or 
declaration, which is probably not what you want
arch/i386/core/hooks.c: In function `arch_main':
arch/i386/core/hooks.c:20: error: structure has no member named `eax'
arch/i386/core/hooks.c:23: error: structure has no member named `eax'
make: *** [bin/hooks.o] Error 1

[tim@server src]$ gcc --v
Reading specs from /usr/lib/gcc/i386-redhat-linux/3.4.3/specs
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man 
--infodir=/usr/share/info --enable-shared --enable-threads=posix 
--disable-checking --with-system-zlib --enable-__cxa_atexit 
--disable-libunwind-exceptions --enable-java-awt=gtk 
--host=i386-redhat-linux
Thread model: posix
gcc version 3.4.3 20050227 (Red Hat 3.4.3-22.fc3)

Regards

Tim


-------------------------------------------------------
This SF.Net email is sponsored by Oracle Space Sweepstakes
Want to be the first software developer in space?
Enter now for the Oracle Space Sweepstakes!
http://ads.osdn.com/?ad_id=7412&alloc_id=16344&op=click