Building SDL2 for Android

Phew is this rather involved.  First you need to download over a gigabytes worth of files. For something that is going to be embedded you need a MASSIVE amount of space.

You need:

I went ahead and installed the JDK in a normal directory,  The same went for the Android SDK.  The NDK is a 7zip file, that I went ahead and put at the root of my C drive because I’m that kind of guy.  Next I unpacked ant into the NDK directory, and SDL.  For SDL be sure to use some command line or other unzip tool that makes sure that the text files are translated appropriately, otherwise they are impossible to read in good editing tools, like notepad.

cd C:\android-ndk-r10d\
unzip -aa \Downloads\SDL2-2.0.3.zip
unzip \Downloads\apache-ant-1.9.4-bin.zip

Now to stage the project directory.  The ‘Android project’ that we are going to build out of is *INSIDE* the SDL archive.  And SDL needs to be inside the project directory, so we xcopy out the bits we need.

md proj
cd proj
xcopy ..\SDL2-2.0.3\android-project\*.* . /e/s
md jni\sdl
xcopy ..\SDL2-2.0.3\*.* jni\sdl /e/s

Now we need to set some environment variables.  Again this is where my stuff is, yours may be different.

set NDK_ROOT=C:\android-ndk-r10d
set ANDROID_HOME=”C:\Program Files (x86)\Android\android-sdk\”
set PATH=%NDK_ROOT%\apache-ant-1.9.4\bin;”C:\Program Files\Java\jdk1.7.0_79\bin”;%ANDROID_HOME%\tools;%ANDROID_HOME%\platform-tools;%PATH%

Now we need to edit the jni\src\Android.mk and point it to your sources.  In this case I’m using dinomage.com‘s simple main.c &  image.bmp as I haven’t written anything exciting yet. Change YourSourceHere.c, to main.c

Now copy in the image.bmp file into the assets directory:

md assets
copy image.bmp assets

Now we can fire up the native compiler tools, and by default it’ll compile for ARM thumb, ARM v7a, and x86

..\ndk-build

Now for the java.  Android is a java platform, so now we have to compile the wrapper that calls our C/C++ code.  First we have to set what version of Android we are compiling for:

android update project –path . –subprojects -t 12

And now we compile with ‘ant’ by typing in:

ant debug

And if all goes well you’ll get:

BUILD SUCCESSFUL
Total time: 2 seconds

Now to run the program, we can fire up one of the emulators (you will have to configure one no doubt) using AVD.  I just clicked around, and picked something that’d launch.  The shell seems to crash a lot, but otherwise yeah.

Now to upload our application to the emulator once it is running:

ant debug install

And if all goes well, you’ll see:

install:
[echo] Installing C:\android-ndk-r10d\proj\bin\SDLActivity-debug.apk onto d
efault emulator or device…
[exec] pkg: /data/local/tmp/SDLActivity-debug.apk
[exec] Success
[exec] rm failed for -f, Read-only file system
[exec] 1985 KB/s (1028625 bytes in 0.506s)

BUILD SUCCESSFUL
Total time: 5 seconds

And on my first shot at running, it crashed.

Unfortunately SDL App has stopped
Unfortunately SDL App has stopped

Using ‘adb logcat’ on the emulator I was able to find this line:

java.lang.UnsatisfiedLinkError: dlopen failed: cannot locate symbol “signal” referenced by “libSDL2.so”…

Well it turns out signal was an inline function until platform android-21, now it’s not inline anymore. When you use the ndk r10, android-21 is used by default but it’s not fully retro-compatible with devices running former Android versions. In this case, signal can’t be found on the emulator. (It did run properly on my Lollipop phone).

Fixing this is simple just edit jni\Application.mk and add in the APP_PLATFORM line:

# Uncomment this if you’re using STL in your project
# See CPLUSPLUS-SUPPORT.html in the NDK documentation for more information
# APP_STL := stlport_static
APP_PLATFORM := android-12
APP_ABI := armeabi armeabi-v7a x86

Re-compile with ndk-build, and re-upload with ant then you are good to go.

working
working

I grabbed, and borrowed some phones around the office, including an x86 phone and yeah it works!

SDL2 test!
SDL2 test!

I haven’t even tried to cross build libraries on Windows… I suspect I should be doing this on OS X or Linux.

Apout on Windows

I read somewhere that kids these days are interested in games where you can modify how the game operates with sub programs written in their own languages etc.

Hello from Unix v7
Hello

So while this does sound interesting, it does remind me of that good old fashioned syscall emulation, where you emulate the CPU, load an executable, but any system calls that are made, are handled by the simulator, little if any hardware is actually emulated. Yes Basilisk II is an example of this, but so is Wine, WOW, NTVDM, i386 code on x64 platforms, and various others. The major advantage is that they typically can access your native file system so you don’t have to mess with virtual disks. Of course it all depends on the implementation.

I did remember this old simulator, Apout, which could run UNIX v6, 7 along with BSD 2.9 stuff on modern Unix. The emulation layer here being LIBC, and how pretty much the basics of how UNIX operates hasn’t changed since those ancient days in the 1970’s.

So I thought I’d try to see how much of this works on Win32 using Microsoft Visual C++ 5. And surprisingly I didn’t have to glue in that much, the biggest thing I had to do was trying to detect if a file about to be opened was ASCII or BINARY as the UNIX platform doesn’t distinguish these two but Win32 with it’s MS-DOS legacy does. As you can see I did get the banner program running, some of the games work ok, although I had to comment out the sgtty functions as there is no immediate equivalent on Windows, and I didn’t think there would be that much of a demand for such a thing anyways. I can even run the login program. Which brings me to the issue which is that it’ll spawn new programs fine, but when an exit is called Apout just exits. Even on Linux it’ll just do this. I tried doing something with setjmp/longjmp but it crashes shortly afterwards… No doubt some stack unwinding fun. As such trying to compile things just bomb out. I went ahead and took the source code to cc and made a native Win32 version that then calls apout for the various parts and that almost worked except I then found out that the assembler on the PDP-11 is a 2 pass assembler, written in assembly. And yes, when as calls as2, and unwinds all hell breaks loose. Which is another problem that UNIX likes to share file descriptors among itself and children, but children like to close things when they die. I guess the solution is to give each child it’s own descriptor table as everyone likes to close stdin/stdout/stderr and even #4, which the simulator uses for logging. it’s very annoying so I just prevented it from closing handles under 4.

But running each of the phases manually does get me an executable but it doesn’t seem to do anything, the only syscalls are closing all the file handles and exiting.

So close!

I don’t think anyone will care, but here is my source/binary along with Unix v7. It’s hard coded to dump stuff into c:\temp for temporary files, the Unix v7 must be in c:\v7 … ugh.

Hunt the wumpus on NT
Hunt the wumpus on NT

But yeah, you can play that thrilling game from 1979, hunt the wumpus!

Did you know that GCC used to support Windows NT ?

I’m not talking about MingGW, Cygwin or anything else like that, but rather when support for Windows NT first appeared, it had to be compiled with Microsoft Visual C++ 1.0 of all things.

It was back in the 2.6 days of GCC, right before the stagnation that led to EGCS.

`i[345]86-*-winnt3.5'
     This version requires a GAS that has not yet been released.  Until
     it is, you can get a prebuilt binary version via anonymous ftp from
     `cs.washington.edu:pub/gnat' or `cs.nyu.edu:pub/gnat'. You must
     also use the Microsoft header files from the Windows NT 3.5 SDK.
     Find these on the CDROM in the `/mstools/h' directory dated
     9/4/94.  You must use a fixed version of Microsoft linker made
     especially for NT 3.5, which is also is available on the NT 3.5
     SDK CDROM.  If you do not have this linker, can you also use the
     linker from Visual C/C++ 1.0 or 2.0.

     Installing GNU CC for NT builds a wrapper linker, called `ld.exe',
     which mimics the behaviour of Unix `ld' in the specification of
     libraries (`-L' and `-l').  `ld.exe' looks for both Unix and
     Microsoft named libraries.  For example, if you specify `-lfoo',
     `ld.exe' will look first for `libfoo.a' and then for `foo.lib'.

     You may install GNU CC for Windows NT in one of two ways,
     depending on whether or not you have a Unix-like shell and various
     Unix-like utilities.

       1. If you do not have a Unix-like shell and few Unix-like
          utilities, you will use a DOS style batch script called
          `configure.bat'.  Invoke it as `configure winnt' from an
          MSDOS console window or from the program manager dialog box.
          `configure.bat' assumes you have already installed and have
          in your path a Unix-like `sed' program which is used to
          create a working `Makefile' from `Makefile.in'.

          `Makefile' uses the Microsoft Nmake program maintenance
          utility and the Visual C/C++ V8.00 compiler to build GNU CC.
          You need only have the utilities `sed' and `touch' to use
          this installation method, which only automatically builds the
          compiler itself.  You must then examine what `fixinc.winnt'
          does, edit the header files by hand and build `libgcc.a'
          manually.

       2. The second type of installation assumes you are running a
          Unix-like shell, have a complete suite of Unix-like utilities
          in your path, and have a previous version of GNU CC already
          installed, either through building it via the above
          installation method or acquiring a pre-built binary.  In this
          case, use the `configure' script in the normal fashion.

Well that is quite the tall order!  I can’t believe I somehow managed to build it back then, and out of curiosity I managed to build it again.  And as you can see you need a ‘beta’ release of the GAS assembler, which is kind of impossible to find now, but it was part of the GNAT project, which used to distribute binary builds of the GCC core back in the day.  Luckily I have a version stuffed away, although it’s from GCC 2.8.1, not as ancient as it would have when GCC 2.6 was a new thing.   But it at lest assembles enough for my simple Hello World experiment.

To build GCC you need a working SED, which surprisingly CL 8.0 from the Win32s SDK, and it runs fine on Windows 7 x64 of all things.  I had to mess with some of the files, and substitute the linker & headder files from Visual C++ 2.0 but much to my amazement not only can I build GCC along with libgcc.a, it also builds incredibly fast.  On my machine I can compile GCC in about 5 seconds.  I remember in 1995 building this on Windows 95 (I was crazy) and it taking HOURS and HOURS.

So one nice thing about these binaries is that you don’t need any external DLL’s  Even the import section of a simple hello world exe is tiny:

GetVersion GetEnvironmentStrings GetCommandLineA ExitProcess RtlUnwind UnhandledExceptionFilter GetModuleFileNameA GetFileType GetStdHandle GetStartupInfoA VirtualFree VirtualAlloc WriteFile WideCharToMultiByte GetLastError FlushFileBuffers SetFilePointer SetStdHandle CloseHandle KERNEL32.dll

Which implies that it only needs KERNEL32 to function.

I don’t know if it’s of any use to anyone else, but HERE is my dump of all the source and tools I used to build GCC.

Shoebill now has working Ethernet support!

Great news!  The excellent A/UX capable emulator Shoebill, now has working Ethernet support!  The sad news is that it only supports the TUN/TAP interface.  So Windows users are kind of left out in the fun.

Shoebill + Ethernet
Shoebill + Ethernet

Except, I’ve been here before with SIMH ages ago.  So I dusted off my source code, and injected it into Shoebill.  The first issue I had was that SLiRP was rejecting all the inputted frames, because of invalid frame length.  Even more weird is that ARP worked, and I could see the 10.0.2.2 and 10.0.2.3 virtual IP’s but TCP and UDP outbound wouldn’t work at all.

It took me longer than it should have but although this code worked great with GCC 2.7 and 3.0, 4.x breaks it.  And it’s the same reason why Shoebill originally didn’t work on Win32, the blasted packed structures!  So adding the ‘-mno-ms-bitfields’ flag to GCC is all it took, and now I could ping 10.0.2.2 for about 5-7 pings until SLiRP would crash.  I tried all kinds of stuff trying to see if there was an issue with SLiRP, but I should have payed closer attention to the debugger, with all those threads flying around.  It turns out Shoebill was trying to read & write a the same time, which caused SLiRP to crash as it is not re-entrant.  I tried to place mutex’s on every SLiRP call but that ended up having SLiRP not process any packets.  Very strange.  I then reduced it to where I read the frame out of SLiRP and pass it to Shoebill, and where Shoebill write’s a frame out the SLiRP.  And much to my amazement I can run ‘worms’ just fine!

So after a minute of worming and pinging I called it ‘good enough’ and rebuilt a production binary, and packaged up my source code.

For anyone who want’s to play, my Win32 EXE is here, and the source code I am using is here.

Fun with PGP 1.0

Well I got slightly bored, and thought I’d dig into some old crypto software.  And PGP 1.0 was as good as any place to start.

Now one scandalous thing at the time was the inclusion of RSAREF 1.0, the RSA reference library which redistribution required a license that wasn’t exactly included with PGP.

A company called Public Key Partners (PKP) holds the exclusive
commercial license to sell and sub-license the RSA public key
cryptosystem. For licensing details on the RSA algorithm, you can
contact Robert Fougner at PKP, at 408/735-6779. The author of this
software implementation of the RSA algorithm is providing this
implementation for educational use only.

And wow was this fun at the time.  As far as I know this license lapsed on September 21, 2000

And then there was this slight issue:

After Biham and Zimmermann go their food and sat down, Zimmermann took out a few pages of computer listings.  Within minutes, Birham was finding fundamental flaws in Bass-O-Matic.  Some of the flaws were subtle-weaknesses that made the algorithm susceptible to differential cryptanalysis, which was Birham’s speciality. Others were more embarrassing, like a conceptual error in Zimmermann’s algorithm that prevented the last bit of each byte from being properly encrypted.  After ten minutes of Birham’s onslaught, Zimmermann realized that Bass-O-Matic was a lost cause.

So now you would be wondering, why would I even bother with what was a quickly abandoned encryption?  Well I was bored, and I was more interested if I could locate the source to 1.0.  What would be more interesting to me is to revive it onto somewhat more modern 32-bit platforms.  Namely OS X, Win32 and MS-DOS.

With a little luck, I found the unix_pgp10.tar.gz, which does contain the source code for a Unix version of PGP!  This version is more so geared to the SPARC of all things. Specifically it mentions:

Tested on SunOS 4.1 with gcc 1.39

However building on OS X was trival with changing the Makefile.  The CC had to be changed to reflect a 32bit build, and the DEFINES had to remove the HIGHFIRST define, as the x86 platform is little endian.

CC=cc -arch i386

DEFINES= -DUNIX -DPSEUDORANDOM -DUNIT16 -DPORTABLE

Is the relevant changes.

And even better it’ll work!

$ pgp pgp.ctx pgp.exe

Pretty Good Privacy 1.0 – RSA public key cryptography for the masses.
(c) Copyright 1990 Philip Zimmermann, Phil’s Pretty Good Software. 5 Jun 91

File has signature. Public key is required to check signature.
File ‘pgp.ctx’ has signature, but with no text.
Text is assumed to be in file ‘pgp.exe’.
.
Good signature from user “Zimmermann, Philip R. – [email protected]”.
Signature made Wed Jun 5 13:51:18 1991

Signature and text are separate. No output file produced.
Plaintext filename: pgp.exe

Wasn’t that great!

Now getting this to run on Windows was a little bit more of a challenge.  I was going to build from the UNIX source code again, however both Visual C++, and Watcom C++ build an executable, but neither are able to add keys to the keyring, verify the executable reliably and deadlock all the time.

So I thought I’d get a little creative and start replacing some code from the MS-DOS version of PGP. It turns out that all I needed was rsaio.c & rsaio.h and I was able to build an executable.  But I ran into other snags, and stack errors.  A glance at the MS-DOS Makefile, and I saw that they had to up the stack size from the defaults.  So I figured the same would hold true, and I picked a much larger 32kb stack for the heck of it.  I mean it is 2014, and if you can’t handle a 32kb stack well..

Compiling on Visual C++ went like this:

cl386 -c /DPSEUDORANDOM /DUNIT16 /DPORTABLE *.c
cl *.obj /F32768 /Fepgp.exe

And for Watcom C++

wcl386 -c -dPSEUDORANDOM -dUNIT16 -dPORTABLE *.c
wcl386 *.obj -fe=pgp.exe -k32768

And now I can build for either compiler.  And even better, it works!

PGP 1.0 on Win32
PGP 1.0 on Win32

Even for completness sakes, DOS4G/W works as well! Just remember to link for MS-DOS

wcl386 *.obj -fe=pgp.exe -k32768 -l=dos4g

And you should be good to go.

PGP for 32bit MS-DOS
PGP for 32bit MS-DOS

So what happened to PGP?  Well version 2 used a more ‘acceptable’ encryption, the IDEA cypher, then the company was sold, IP was sold again and again.  It’s still out there, mostly for email encryption.

While it sure did ignite the world on fire for a while, the overall difficulty of using it, combined with the ease of losing the private key and all your data is just too easy.  But this really is the nature of the beast.

Shoebill ported to Windows!

Shoebill!
Shoebill!

Good news, as mentioned here, the Shoebill emulator was recently given some much needed SDL love, and ported to Linux.

Well that’s great and all, but the vast majority of people who run anything these days do it with Windows.  So I decided to try to get it to compile with MinGW to see how far I could get.

And the short version is that I got it working!

The long version is that in the first pass there is some SIGUSR2 stuff that is undefined.  And for a good reason, since it won’t work.  So I just commented them out.  The next minor problem was the lack of bzero.  Honestly I don’t know why bzero is missing from MinGW, but who knows why.

Shoebill also processes some internal macros with a perl script that for some reason was dropping in binary values into the source, making GCC mad.  I just commented out a line that was adding in more comments into the header.  This let me compile with a simple pass.

There was some issues reading the ROM file, since the 68000 is a BIG ENDIAN processor, and the 8086 is LITTLE ENDIAN, Shoebill makes extensive use of hotns and hotnl, ntohl, and ntohll.  These can be found in the winsock library, and even better they dont need any winsock initialization, they work right away.  I just have to make sure I include winsock2.h, and link against the winsock library.

However when trying to boot, the checksum was 0x00000000, not the expected value!   Luckily there was an assert to catch that and crash.  This led me to notice that in Linux files are opened in binary mode by default, while on Windows, they are opened in ASCII mode.  A quick change of all the fopen calls, and I was reading the ROM, but now crashing on the disk.

As it turns out newer versions of GCC go all crazy when it comes to structs, and try to automatically align to boundaries for quick access.  Which sound nice, until you try to read in some binary data, and expect things to be in certain locations and find out that your structure is larger than expected, and data is read in the wrong place.

The solution is to force the compiler to leave it alone with

__attribute__ ((__packed__))

HOWEVER as luck would have it, Microsoft apparently packs structures a different way, and you have to either make a macro to do a bunch of work to force it to make the structure 1:1 of what you expect, or use the CFLAG option of

-mno-ms-bitfields

And now MinGW’s GCC will build something along the lines of what it’d build on Linux.

Putting it all together, I amazingly got this!

Shoebill on Windows
Shoebill on Windows

Phew!  So for those interested, here is the source code drop(Use the updated one here!), and here is the binary.

If you ever wanted to see the “OS X” of the 1980’s, now is your chance!

Qemu 1.6.0 for Win32

I updated my glib2 to the same version that works great on OS X, and it feels like the Win32 binaries are just slower than ever.  Now I am running this under VMWare, so maybe that is why it is so slow.

Can someone give this a shot?

Any and all feedback would be nice.

Oh I included the Doom 1.1 in there run it like this:

qemu-system-i386 -L pc-bios -m 16 -hda doom11.img -soundhw sb16,adlib

And you should be able to run doom with sound.

I’ve removed the fullscreen/resizing on Qemu 1.1.1

And now it is *MUCH* faster…

Next I’ll have to investigate better audio libraries, as winwave seems to leave a lot to be desired.

Basically the changes in ui/sdl.c were:


case 0x21: /* 'f' key on US keyboard */
// toggle_full_screen(ds);
// gui_keysym = 1;
break;
case 0x16: /* 'u' key on US keyboard */
// if (scaling_active) {
// scaling_active = 0;
// sdl_resize(ds);
vga_hw_invalidate();
vga_hw_update();
// }
// gui_keysym = 1;
break;

And


case SDL_VIDEORESIZE:
// sdl_scale(ds, ev->resize.w, ev->resize.h);
vga_hw_invalidate();
vga_hw_update();
break;

So now it keeps the correct video display size, so it isn’t going like crazy trying to scale the screen 30 times a second..

Doom now looks normal!

As you can see stuff like Doom now looks normal.  As mode changes are initiated by the video card, it keeps the scale to where it was in prior versions.  At least its not going 1:1 native as looking at a 320×200 window on a 1280×768 desk would be a tad hard..

And yes, even things like Windows 3.0 look correct when the screen changes resolution:

Windows 3.0 in 386 enhanced mode

Also I should add that if you are going to try to use disk images that are *NOT* 1.44 MB they will not work.  You’ll have add this flag to Qemu:

-global isa-fdc.check_media_rate=off

I’ve been told the new handling of disks is better in this version so I’ve left this setting where it was..

I have just updated the download link, but for those who missed it, you can download the i386  win32 version here.

Building Qemu 1.1.1-1 for Win32

I still can’t see how to build this for win64 … 🙁

The first major stumbling block on Win32 is glib, and again going back to the MinGW wiki there is a good laundry list of how to bootstrap Glib on MinGW.

Things you’ll need:

Be sure to run ming-get install for the following:

  • gcc
  • g++
  • libiconv
  • zlib
  • libz
  • gettext
  • msys
  • msys-perl
  • msys-m4

Install Python.. Nothing to fancy from what I gather Qemu still requires you have a level 2 Python, not 3 …

Build libffi

This should be the usual dance of configure & make / make install…  But sure to at least run configure like this:

./configure –prefix=/mingw

Build Glib without pkg-config

Be sure to add Python to your path..

PATH=/c/Python27:/c/Python27/DLLs:$PATH
export PATH

Glib is packed with something called ‘xz’ so hopefully you have xzcat .. Otherwise add it!

export LIBFFI_CFLAGS=’-I /mingw/lib/libffi-3.0.9/include’
export LIBFFI_LIBS=-lffi
export lt_cv_deplibs_check_method=”pass_all”
export CFLAGS=”-O0 -g -pipe -Wall -march=i486 -mms-bitfields -mthreads”
export CPPFLAGS=”-DG_ATOMIC_OP_USE_GCC_BUILTINS=1”
export LDFLAGS=”-Wl,–enable-auto-image-base”
configure –prefix=/mingw –with-pcre=internal –disable-static –disable-gtk-doc –enable-silent-rules

You may have some weird issue where when running configure it tells you it cannot create executables, or you get a bunch of weird errors trying to paste in the CFLAGS line.. For me the MinGW prompt was stripping the quotes and the leading – to the -O0 (disable optimization) bit.  I don’t know what on earth its issue was, but I had to type that line in manually.

Then do the make/make install dance. This will take a WHILE. At least with -pipe it’ll run each stage of GCC on multiple processors… But yeah.. This is intense to build.  Good thing we get to do it twice.

I also ran into some weird error in the GIO directory where it couldn’t find my Python, and was looking for python2.5 .. So I copied python.exe to python2.5.exe …

PYTHON = /usr/bin/env python2.5

pkg-config

Naturally pkg-config depends on Glib2, and pkg-config to build… Which of course is a circular problem, much like Glib2 requires pkg-config to build.   So to configure it, it goes something like this now that we’ve built a Glib2 ..

export GLIB_CFLAGS=”-I/mingw/include/glib-2.0 -I/mingw/lib/glib-2.0/include”
export GLIB_LIBS=”-lglib-2.0″
configure –prefix=/mingw

At the same time I can’t help but wonder if this version of pkg-config can use it’s own Glib with the following config:

configure –prefix=/mingw –with-internal-glib

Anyways with any luck we can now build & install pkg-config.  This only takes a few seconds..

Glib2 (again)

From this point you should open a new MinGW prompt window, as you don’t want the old CFLAGS screwing things up.  Re-eport the Python path/dll vars and now we can get on to building glib2 (again!) …

configure –prefix=/mingw;make;make install

 

SDL

This should be somewhat straightforward …

configure –prefix=/mingw;make;make install

In the old days we built zlib, but now we can just quickly add in the package (as we did way above) so we should.. now be ready for the main event!

 

And now we can finally build Qemu 1.1.1-1!!!

Now there is a few things I like to tweak, the first is in the configure script, I like to add in AdLib support.  Look for the line

audio_card_list=”ac97 es1370 sb16 hda”

and add adlib into the list

audio_card_list=”ac97 es1370 sb16 hda adlib”

Next I like to modify hw/pc.c and alter the ISA NE2000, as Qemu doesn’t like to share IRQ 9 with the card, so it is just easier to remove the 0x300/IRQ 9 definition.

static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
0x280, 0x380 };
static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };

to this:

static const int ne2000_io[NE2000_NB_MAX] = { 0x320, 0x340, 0x360,
0x280, 0x380 };
static const int ne2000_irq[NE2000_NB_MAX] = { 10, 11, 3, 4, 5 };

The next tweak deals with the ability to use older qcow2 disk images… I guess converting them to RAW with an older version of Qemu, then using the new version of Qemu to convert them back into qcow2’s may be a “good idea™” but for now modifying the source is a quicker fix.

Comment out the “return -EINVAL;”  in block/qcow2.c

if (ext.len > end_offset – offset) {
error_report(“Header extension too large”);
//return -EINVAL;
}

Now one fun thing I’ve noticed is that building with the default O2 flags Qemu will crash out the moment you access a hard disk image.  It appears that coroutine-win32 is at issue (again?).  So the “easy” way I address it is to first build qemu as normal, and verify that if you attach a hard disk image (any kind) and try to access it, partition it etc, it should crash.  Next remove the file coroutine-win32.o , and edit the file config-host.mak and change the CFLAGS that specify

CFLAGS=-O2 -g

to

CFLAGS=-O1 -g

Now run make again, and it *should* just rebuild coroutine-win32.o with the lesser optimization flags, and relink all the exe’s.  If I’ve done this right, you should now have a working Qemu.

You can go ahead and strip the binaries if you so please, but that should be it.

PHEW.  For anyone who wants my build, but doesn’t want to go through this ‘exciting’ process, you can find the Win32 i386 build here.

Weird scaling in action .. Control+ALT+u kind of undoes it, but it just doesn’t look right and it is far too slow.

In preliminary testing I’ve found this version to be MUCH slower than 0.15.1 .. I think it has something to do with it wanting to scale the SDL window.  Also I’ve had issues with various network cards not initializing with the BIOS that ships with this version of Qemu so I’ve included the bios directory from 0.15.1 .  And lastly yes the disk images… I’ve had major issues with my qcow2 disks, and disk corruption with this build.  I’ve gone ahead and  included the qemu-img tool from 0.15.1 so you can convert qcow2 to raw, then use the 1.1.1 tool to take them from raw back into qcow2 … But I’d probably only do it as a test.

You may want to kick the tires on this version but 0.15.1 really blows this one away…

Running Qemu as a Windows service …

Long story short I’m doing some work with a network that suffers a lot of ‘you can’t get there from here’.  They’ve given me VPN access and yet even the VPN cannot get to a lot of stuff.

The solution for them is to use this old server and ssh out from there to the rest of everything.  Which for the most part works fine, but if more then 2 people need to leapfrog suddenly you are waiting in line, or constantly knocking people off.

So I figured I’d do something different, install a QEMU virtual machine on the server during my allotted hour, and then launch it as a service so that I could leap in/out through the VM leaving the console free.

While I am going to add Qemu as a service, it is still somewhat stealthy as I don’t need device drivers, and I can run it nested as I know this machine is slated to be migrated to VMWare ESX.  And the best part of that is that it’ll continue to run.

So how do we set this kind of thing up?

The first thing you’ll need is srvany.exe instsrv.exe which both can be found in the Windows 2003 resource kit.

Installation is very straightforward, just remember to use complete paths for your Qemu, BIOS, and disk files.  Installation goes like this from the command line:

InstSrv qemu c:\qemu-0.15.0\srvany.exe

This will create a service entry named Qemu, which will in turn kick off the srvany executable from the resource kit.  Now I know what you are thinking, what about Qemu?  Well we have to specify that using regedit.  Also remember that because you are going to run this as a service you don’t want the SDL display popping up and scaring some poor hapless user.  So the first thing I’d recommend is to work out the flags that you want to start with.  Something like this:

c:\qemu-0.15.0\qemu -L c:\qemu-0.15.0\qemu\pc-bios -hda mydisk -net nic -net user -redir tcp:2222::22 -vnc w.x.y.z:2223

This will redirect tcp port 2222 into the VM for ssh, and sits the VNC display on port 2223 …

So we fire up regedit and navagate to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Qemu

There we add a new key “Parameters”.  Then add an ASCII key of “Application” then just paste in the all of the qemu flags as mentioned above (or changed as needed by you).

Then you can simply start/stop the thing using the net start/net stop.

I suppose this is a little subversive (lol) but sometimes you’ve got work to do and the best way through it to piggyback on someone else’s computer.  Also I really fail to see the ‘wisdom’ in creating ACLs that only permit you to access your routers/switches from your desktop when you could easily *NOT* be in the office.  Or this guy just likes the excuse of not being able to work from home.

Anyways not to ramble but that’s how I ‘fixed’ the issue without ruffling too many feathers.