- The 'Bits' type used in the opl files is incompatible with Bit64s. Modified
typedef to fix this issue.
- On Windows use Sleep() directly, since msleep() doesn't exist in Cygwin.
USB UHCI and the E1000 NIC are known to work fine on some guest systems, so
it's safe to remove this attribute.
TODO: Some other Bochs features need to be reviewed for this.
- the "remaplist" has 256 entries, but the limit was too high (4096).
- the "patchtable" had 4096 entries, but the index was never > 15.
- removed disabled / outdated code.
- Undocumented "legacy" register at offset 0x1b is used to probe for the ES1370
interrupt by the Win9x driver. Using bit #0 to control the IRQ line fixes
driver installation.
- Added more detailed BX_ERROR messages for the UART registers.
- Small code cleanup in the write handler.
- TODO: fix possible bluescreen in Win98 guest after wave playback.
- OPL3: apply FM volume control to the output data.
- SB16: fixed missing volume update after mixer reset.
- soundlow: fixed application of the volume control to the PCM output.
- Removed useless set_pcm_params() call in base class contructor.
- SDL waveout destructor needs to call the unregister_wave_callback() method
of the SDL stuff, since the base class destructor doesn't do that.
- Moved BX_INIT_MUTEX() call to start_mixer_thread() to reduce code duplication.
- Call BX_FINI_MUTEX() only when the mixer thread and mutex was really used.
NULL for all services and the dummy driver provides the stubs. Now falling
back to the dummy driver if the wavein or midiout services are not provided
by the selected driver.
- sound 'file' driver: added failure handling in the file open methods.
- soundmod: initialize default driver with NULL to avoid crash on failure.
- SB16: Fixed crash if no driver is present / removed unused members.
- Implemented wave output using the 'soundfile' driver including dual wave
output in wavemode 3. Removed now obsolete output file handling and other
cleanups. The wave input (recording) feature is not affected by the
wavemode setting.
- Moved VOC file handling to the soundfile code.
- Updated iodev makefile dependencies.
- Implemented wave and midi output using the 'soundfile' driver.
- Implemented dual midi/wave output in midimode/wavemode 3.
- Removed now obsolete output file handling and other cleanups.
- The wave input (recording) and OPL3 output features are not affected by
the wavemode setting.
- TODO: Similar rewrite of ES1370 wave output and related cleanups.
- soundlow: use direct output if no PCM callback is registered.
- soundlow/soundfile: re-implemented openwaveoutput() method to support
runtime changes of wave mode or file (only implemented in the 'file'
driver, stub for others).
- soundfile: don't initialize mixer thread and PCM callback if thread is
already running / no output() delay in that case.
- Added the capability to load up to 4 sound drivers and added methods to
get pointers to the waveout, wavein and midiout services of a driver.
- Rewrite of the SB16 midi file output code. In midimode 2 and 3 the midiout
service of the 'file' driver is used. Removed obsolete midi file handling.
- Re-implemented the closemidioutput() method to make runtime changes of the
midi setup work with the new code.
- waveout: added support for RIFF WAVE format and raw output. The formats
'voc' and 'wav' are detected by file extension, raw output is used otherwise.
- midiout: added support for raw midi output (no header, no deltatime).
The 'mid' format is detected by file extension, raw output is used otherwise.
- Added 'midiout' parameter to the global sound option to specify the device
if necessary. The related SB16 parameter now only specifies the output file.
- TODO: Rewrite of the sound configuration and output code to support dual
output (device and file).
- New code is based on the dummy sound driver and file output code from SB16.
- The midiout feature supports the MIDI file format.
- The waveout feature currently only supports the VOC format.
TODO:
- Add support for a more detailed sound configuration (waveout to device,
midiout to a file.
- The waveout feature should support WAV and RAW file formats.
- Move VOC file handling from soundmod.cc to soundfile.cc.
- Finally remove file handling code from sound devices.
creates the object and returns a pointer to it. TODO: Rewrite of the sound
configuration to support a more detailed setup (e.g. waveout with alsa,
dummy wavein, midiout to a file).
creates the object and returns a pointer to it. TODO: The wavein and midiout
components should also move to separate classes. When finished, it should be
possible to configure the sound support more detailed (e.g. waveout with sdl,
midiout with alsa and the dummy driver for wavein). Creating a separate
module for the output to file(s) is also possible.
'alsa'. Cleaned up the code a little bit.
TODO #1: Implement mixer thread support in the OSX sound driver.
TODO #2: Code cleanups in the Bochs sound code.
TODO #3: Resampling support and improvements in the wave recording code.
- removed disabled code for the sndPlaySound() function.
- removed now obsolete buffer ring for wave output.
- The new mothod waveout() is driven by the mixer thread. For the best results
you need to disable the 'realtime' sychronization and to find out a usable
IPS vaöue. The mixer also polls data from the speaker beep generator and
the OPL3 FM generator.
- TODO: Code cleanup in soundwin.cc, implementation in other sound modules.
- soundlow: prepared code for 2 pcm parameter sets:
emulated: requested by sendwavepacket()
real: used for the sound output
- soundsdl: startwaveplayback() rewritten to set_pcm_data().
- soundmod: some beep init code fixes.
- OPL3: Added support to change the samplerate of the generator.
- SDL sound: convert wave packet to 16 Bit signed stereo format.
- Increased the maximum size of a single wave packet for playbck/recording.
- Beep thread now only started one per simulation.
- Added some more multithreading macros and use them in rfb, vncsrv and soundmod.
- TODO: implement mixer thread and remove beep thread.
The code currently only works with the SDL sound output module and at 44100 Hz,
with 16 bit stereo signed little endian format. FM volume control and
save/restore support are not implemented yet. The legacy code for generating
MIDI notes still works and will be removed when mixing support is implemented
in all sound modules.
- Define new structure bx_pcm_param_t and use it as a parameter for the method
sendwavepacket(). The method startwavepacket() is now called from
sendwavepacket() in case of a parameter change (temporary solution).
- soundsdl: Replaced waveready() call by some more accurate code for buffer
space check in sendwavepacket().
- soundsdl: Moved code for reading from audio buffer to a new function.
- soundsdl: Temporarily increase the audio buffer size.
- Preparing the SB16 code for the OPL emulation port from DOSBox.
- Reworked wave callback code to support 2 input sources (no mixing yet).
- Fixed SDL sound output.
- Beep generator fixes.
- Prepared the SB16 code for FM OPL output support (to be ported from DOSBox).
- Renamed wavedata callback type.
- Handle sample rate in the beep generator (TODO: handle data format).
- Don't call waveready() in the SB16 code.
- Attempt to fix SDL2 wave output using SDL_LockAudio() / SDL_UnlockAudio().
- SDL specific code in the beep thread no longer needed.
- Added framework for polling wave data and use it for beep output in sdl.
- Some work in the beep thread code (to be continued).
- Moved macros required for multithreading to bochs.h.
- Increment timer counter before checking for overflow.
- Implemented ISA bus delay in the pc_system code and use it for the SB16.
This makes OPL chip detection work at ips values > 4 Mips.
TODO: OPL chip should use wave output instead of MIDI (maybe port from DOSBox).
TODO: Use isa_bus_delay() for other legacy devices if required.
- feature implemented in the wxWidgets gui
- feature not required for legacy SDL (releases keys on focus change)
- removed duplicate code in keyboard.cc
- TODO: implement feature in the win32 gui if required
The timers now have a new member 'realtime' and they are driven by the
selected engine. The VGA update timer and the status LED timer now always use
the realtime mode, but the PIT and CMOS RTC depend on the clock options.
TODO: improve virtual timer to make operate in both modes at the same time.
The vga update code should always use the realtime mode and the pit timing
should depend on the clock option.
Notes by Ben:
This emulates a NEC/Renesas uPD720202 2-port (2 socket, 4-port register sets)
Root Hub xHCI Host Controller.
Many, many thanks to Renesas for their work and effort in helping my research.
I have tested in with my own tests and WinXP Home Edition SP3.
Use port1 and port2 to emulate a Super-speed device, and use port3 and port4
to emulate Low-, Full-, or High-speed devices.
NOTE: The Windows driver for this device uses a proprietary detection
mechanism and we don't have the permission to implement it. That's why it only
works fine with non-Windows guests.
- removed max_packet_size member from USB device structure
- usb_uhci: improved speed info
- updated USB-related documentation
- TODO: review and test the main part of the xHCI changes by Ben
the init code. Started with the SeeReg[] array values. Some other setting could
be added later. The parser code is based on the slirp config parser.
Slirp: fixed possible segfault if the value for a parameter is missing.
- added new symbol BX_WITH_SDL2 and prepared build system
- TODO #1: add new option --with-sdl2 to configure script
- TODO #2: add new file sdl2.cc (porting seems hard to do - writing from scratch
might be better)
- embedding combined manifest makes BOCHS.EXE usable
- list of DLL plugins now depends on the pci and debugger options
- 'make clean' now removes the generated manifest files
- renamed variable for external rules
- documentation update
complete and usable with the BOCHS.EXE built with the IDE.
TODO #1: fix BOCHS.EXE built with nmake (manifest issue)
TODO #2: make the currently fixed DLL plugin list variable
sections. A better solution must be found for the basic set of device plugins
in iodev that don't require special rules.
Switched back to the original gui macro naming for linking plugins.
rules can be built now (keyboard, pit, gameport, serial, vga, svga_cirrus,
hdimage, netmod, soundmod, usb_common). The generated DLLs are usable with the
BOCHS.EXE built with the IDE, but the nmake won't work at all yet.
(bx_hdimage.dll) can be built now. All plugins that require special rules
could be fixed for nmake in a similar way.
TODO: find a way to compile plugins with standard rules (GNU make uses bx_%.dll
in that case, but MSVC nmake doesn't support this syntax).
required an external binary and it couldn't be ported to Windows. Now the new
builtin and portable slirp support based on Qemu's implementation is almost
stable and all the 'vnet' DHCP and TFTP features have been ported to the new
slirp module.
TODO: rename all the "slirp_new" stuff to "slirp"
- removed obsolete local variables that made the hostname and bootfile options fail
- store requested hostname as a string to simplify the code
- check free space while processing the parameter list
- setting the max. DHCP message size is not supported yet (send error to log)
TODO #1: implement all features of the existing (vnet) TFTP support in Slirp.
TODO #2: make the slirp DHCP server work like the vnet one.
TODO #3: finally remove the slirp backend and rename the new one to "slirp".
- bx_param_string_c method isempty() didn't work for raw byte strings. Now using
it to check whether or not the current value matches the initial one. The
parameter handling of the network adapters now use this method to check if
the MAC address is already initialized to avoid incorrect BX_ERRORs.
- small documentation update
TODO: 'dnssearch' option should accept a comma-separated list of suffixes.
Other slirp TODOs: SMB support on Linux, reduce code duplication (ARP, DHCP, TFTP)
- added support for a separate slirp config file specified with the "script"
parameter. Added sample config in the "misc" subdirectory.
- now creating one logfunctions object per slirp instance and added new function
slirp_warning() to send error messages to the Bochs log.
- regenerated dependencies for the iodev/network Makefile and updated command
line example for generating them.
Updated workspace files after adding slirp and avx sources
Updated slirp TODO list:
- reduce code duplication with 'vnet' module (DHCP, TFTP, ARP)
- add separate config file for slirp to make it more flexible (using the
'script' parameter)
- add SMB support on Linux
- remove slirp backend module after next Bochs release
- applied remaining updates from Qemu slirp
- add #if blocks to all slirp .cc files for conditional compilation
- fixed some uninitialized variable warnings
- added container_of() macro replacement
- fixed unresolved symbol in netmod.cc
- added MSVC-style structure packing pragmas
- check for the presence of min / max macros
- enabled slirp_new for MSVC in configure script
TODO list:
- check for more updates from Qemu
- fix lots of compiler warnings
- reduce code duplication with 'vnet' module (DHCP, TFTP, ARP)
- add separate config file for slirp to make it more flexible (using the
'script' parameter)
- update MSVC workspace files
- add SMB support on Linux
- remove slirp backend module after next Bochs release
- rewrite of the logical sector handling code to make seek latency support possible
- implemented variable HD seek latency for seek and read commands
- TODO #1: CD-ROM seek command returns status immediately, but it clears the DSC bit
and sets it at completion. Device should not accept ATAPI commands until DSC is set.
- TODO #2: clean up the seek latency code
- temporarily disabled seek latency support for ATAPI command 0x2b until the
seeking status is implemented correctly
- fix for the MSVC compilation fix (result was always 0.0)
is almost a port of the Qemu 1.1.2 implementation. It has been tested successfully
on Linux and Windows (MinGW/MSYS). The networking module is currently called
"slirp_new", since the "slirp" backend module still exists.
TODO list:
- MSVC support (requires container_of() macro replacement)
- apply fixes and improvements from recent Qemu releases
- use our TFTP implementation to reduce code duplication
- check if we can share ARP and DHCP support with "vnet"
- add SMB support on Linux
- finally remove the slirp backend module support
- now using the more accurate name "seek latency"
- fixed CD-ROM LBA address range checks and renamed limit to "max_lba"
- make the latency for CD-ROM access variable depending on the distance between
current and new block address, Maximum value should only occur at first access
or after media change.
- TODO: make hard disk seek timing variable, too.
- still TODO: seek latency for USB drives (requires asynchronus packet support).
- save function implemented using a dummy parameter that only receives the
result of the save operation. The parameter's save/restore handlers doing
the main job. The numerical values of the requests are saved in a separate
text file. The data transfer buffers are saved as binary files.
- the request data buffers are now created dynamicly
- TODO: restore SCSI requests
A device can set it with setTimerParam() and read it in the timer handled using
triggeredTimerParam(). Simplified some timer handler by using these methods.
TODO: implement HD / CD seek timer
- fixed clock values (assigned to wrong model)
- LFB read/write: don't limit y values to 1023 (4MB now usable this way)
- additional Voodoo2 fbiInit* register writes now handled like the Voodoo1 ones
usb device plugins. Loading the plugin means that the user wants to enable it.
Disabling the device on the command line and in the config interface is still
possible.
- destructor: free memory only when initialized
- parameter 'enabled' must be initialized with 1 (TODO: fix other plugins)
- PCI register fixes (Voodoo2 now detected, but not usable)
- fixed pluginlog init in non-plugin mode
- reduced logprefix length limit and use it for the related string parameter
- increase the device prefix size to 6
- TODO: change some more device prefixes
The device exists for some years now, but it was never functional and it is not
yet mentioned in the docs. After studying Linux kernel sources and a lot of
tests it is now usable. A configure option doesn't exist, so it must be enabled
in config.h. At runtime it is only loaded if the mouse type is set to 'bus'.
that is a part of the libtool package. Updated Makefile dependencies.
TODO: check if we can get rid of the ltdl*.* files (this would be possible if
the ltdl library is always available if libtool is present).
- ES1370: implemented CODEC registers (AK4531 mixer) and wave volume control
- Moved wave volume application code to new method pcm_apply_volume() and use
it for both soundcards
- TODO: stereo support
wx floppy dialog now forces a media change by clicking OK (same as win32dialog)
wxdialog.cc: one more whitespace cleanup
wxdialog.h: removed unused members
- es1370: fixed wave file parameter name
- wx cdrom dialog: use the first item of the cdrom runtime parameter list
- wx ParamDialog: always pause the simulation thread while setting parameters
- wx ParamDialog: call SIM->update_runtime_options() before continuing simulation
- SB16: log file name now also available as a runtime option
- SB16: changed order of the runtime options and minor other changes
- ES1370: changed the "wavemode" parameter type to bx_param_enum_c and set up
more accurate dependencies (similar to SB16)
- ES1370: wave mode and file name now available as a runtime option
- If the runtime flag of a bx_list_c object is set, all it's items are set up
as runtime options.
- Set the runtime parameter for all top-level runtime-only lists. Setting up
each single parameter is no longer needed.
- Always initialize the lowlevel sound driver (for wave input / recording)
- Wave input (recording) is now independent from the wave output mode
- Changed the "wavemode" parameter type to bx_param_enum_c and set up more
accurate dependencies
- Minor related changes
- use #if BX_SUPPORT_CDROM instead of #ifdef LOWLEVEL_CDROM
- devices with cdrom support now use the base class cdrom_base_c
- cdrom count variable moved to cdrom.cc
- cdrom_misc.cc: don't use ioctl's for ISO files
In convert/resize mode a backup can be created if the new file name
is not specified or the same as the source file name. In commit mode
both base image and redolog file are affected by the backup switch.
Added Linux specific file copy code using '/bin/cp'.
- added support to use 'concat' and 'dll' mode images as convert source.
A prefix ("concat:" or "dll:") must be added to the path since autodetection
doesn't work for these modes.
- create: fixed virtual disk size calculation based on C/H/S
- convert / resize (interactive mode): use original filename as the template for
the new one if not specified on the command line
- added support to resize images (currently supported: flat, growing and sparse)
- added image file name check and renaming if necessary
- set minimum hard disk size to 10 MB
- vmware4: make header structure available for usage in bximage
creation, conversion and rodolog commit. Image creation support is not
present yet and the executable has not been added to standard compile yet.
TODO #1: add floppy and hard disk image creation support (from bximage.c)
TODO #2: implement support for conversion to vmware and vpc formats
TODO #3: add MSVC (nmake and workspace) support
Detect and reply all supported TFTP extension options for both read and write
operations. Use 'blksize' value if smaller than buffer size.
TODO: increase buffer size, implement timeout mechanism
- ALSA driver is now default on Linux (if present)
- SDL sound driver is now default if no platform-specific driver is available
and the SDL gui is enabled
- soundlnx.cc: sendwavepacket() now exits with error if device is not opened
- soundmod.cc: exit beep loop if sendwavepacket() returns error
- ALSA stuff moved from soundlnx.* to new soundalsa.* files
- SB16 / ES1370: including specific sound driver headers not necessary
- updated dependencies
- added parameters 'wavemode' and 'wavefile' to the 'es1370' option
- ES1370: added new method sendwavepacket() (similar to SB16)
- moved VOC file methods from sb16.cc to soundmod.cc and added macros for them
- SB16: minor fixes
- added 'wavein' parameter to specify the wave input device if necessary
- if 'wavein' is not set, Bochs uses the 'waveout' value for input, too
- SB16: the 'wave' parameter is now only used to specify the filename (mode 2 or 3)
- TODO: implement 'wavemode' feature for ES1370 similar to SB16
- added 'sound' option to select the driver and the wave output device
(see bochsrc sample for details)
- the wave device option of sb16 and es1370 is now only used to set up the
wave input device (if necessary)
- cleaned up the speaker code to initialize the 'sound' case correctly
- memory: initialize ISA and PCI ROM variables in init_memory()
- sb16: remove items from runtime menu in destructor
- gui debugger: show dialog contents (TODO: fix remaining display issues)
- TODO: fix deadlock in bx_wx_gui::handle_events()
- bx_list_c: added support for removing parameter with other parent list
- USB HC's: remove item from runtime config on simulation exit
- initialize bx_user_quit in bx_begin_simulation()
- only remove optional plugin from list if alrady unloaded
- isempty() returns 1 if string length is 0 or string has special value "none"
- sprint() copies the formatted / converted string into a buffer
- use new methods in some parts of the code (TODO: win32 dialogs)
- check for the special value "none" to avoid failures
- removed parameter "present" and use the new drive type "none" instead
- re-implemented cdrom media status parameter as a bx_param_enum_c
- TODO: change floppy media status parameter type to be consistent
- fixed bochsrc line after a "raw byte" string parameter
- changed parameter name "macaddr" to "mac" (same as bochsrc option parameter)
- skip disabled parameters when building the bochsrc line
- added "multiline" switch to select the output format (one or multiple lines)
- renamed siminterface method and use it in the network devices code
- renamed config parameter "i440fx_support" to "enabled"
- new config parameter "chipset" added (current choices "i430FX" and "i440FX")
- don't load ACPI support if the i430FX chipset is selected
- select register values for the core PCI devices depending on the chipset
- USB UHCI must be connected to a PCI slot if the i430FX chipset is used
- rombios changes to make the i430FX chipset work
- TODO #1: implement limitation to 1 cpu and 128 MB RAM for the i430FX chipset
- TODO #2: verify register behaviour of both chipsets
- read PCI confAddr value from the devices code for debugger output
- PCI confData value isn't needed at all (changes by each read / write access)
- PCI framework methods don't need to be virtual
- PCI slot configuration, memory and i/o BAR registration, PCI i/o space
handlers now present in devices.cc
- configure: E1000 emulation and USB support need PCI
- vga: fixes for the non-PCI case
- renamed trigger_timer() methods to refresh_display()
- new argument 'redraw' added to force a refresh of the entire screen
(used in snapshot mode)
- siminterface method refresh_vga() now checks if the Bochs gui init is complete
- use SIM->refresh_vga() in some more cases
- graphics_tile_info() split into a common and specific part
- store snapshot mode in structure bx_svga_tileinfo_t
- in snapshot mode, display adapters copy raw data to buffer
- rewrite of the trigger_timer() code to make it work in all cases
- added new class bx_nonvga_device_c to make forwarding refresh requests to the
Voodoo adapter possible
API) moved to the gui code. Splitted graphics_tile_update() in a common and a
specific part. The common part handles the snapshot case if active. As a side
effect, the snapshot feature now works for CGA modes, too. Fixed palette colors
in 8 bpp snapshots.
the display adapter code. When ready, the common gui code should handle it
completely. The snapshot handler should set a flag 'snapshot_mode' and then it
should force a display update. The graphics update code should be redirected to
some code that 'draws' to a snapshot buffer. These first steps are done:
- store some guest display settings in the gui code (text / graphics switch,
x / y resolution and bpp)
- split 8 bpp palette update code in a common and a specific part and store the
palette values in the gui code
These hdimage modes are now supported: flat, sparse, growing, vmware3, vmware4
und vpc. The image mode is auto-detected, so no change of configuration syntax
is necessary (TODO: documentation update). Example:
ata0-master: type=disk, mode=volatile, path=vmware4-test.img
- redolog class: added support for opening redolog file in read-only mode
- open backups for header check in read-only mode (growing, undoable, volatile, vpc)
type of the r/o disk for the undoable / volatile mode. The format check code
is currently written for flat, sparse and growing images. The undoable and
volatile mode now cause a panic if the r/o image mode is unsupported.
- TODO #1: rewrite the header checks for vmware3, vmware4 and vpc images to make
them usable for autodetection.
- TODO #2: add support for other types of r/o disks in undoable/volatile mode.
- TODO #3: add an image mode 'auto' for the supported types (NOTE: unknown images
can be treated as flat ones if their size is multiple of 512).
- don't enable Voodoo mode if hSync or vSync are still 0 to avoid Bochs crash
(division by zero error)
- handle PCI register 0x40 (initEnable)
- fixed reported display size
time if requested (reduces code duplication)
- 'vpc' image mode: moved header check to a separate function that opens the
file, checks header/footer and returns disk type (if requested)
- implemented restore function for 'vpc' mode images
- gui: updated yes/no dialog message after saving state
- save/restore documentation update (TODO: developer doc)
be set for "true", but variable "lfb_enabled" wasn't correctly set up.
- TODO: check the whole Bochs code for the correct usage of variables saved
with bx_param_bool_c.
