There are cases where when it's necessary (e.g. given uninitialized NVRAM,
the Beige G3 with the 10.2 install CD inserted will update the boot
device and restart to boot from it).
Restart support was done by wrapping the ppc_exec function in a loop and
checking for a restart power off reason. We also need to disconnect all
event listeners, since they will be recreated when the machine is
re-initialized.
When fetching DMA command, make sure to convert little endian fields to host endianness (i.e. don't use memcpy).
When fetching DMA command, return the host address of the DMA command for LOAD_QUAD. Maybe this address should be cached whenever this->cmdptr changes?
When fetching DMA command, return whether the DMA command is writable.
For LOAD_QUAD, pass the address of the DMA command to xfer_quad.
Always log unexpected DMA command values.
Write full 32-bit value for LOAD_QUAD.
Write reqCount to resCount.
pci_conv_rd_data can be used to handle unaligned or 64-bit accesses in mmio regions if it's modified to include the next 32-bit value.
For pci config accesses, grackle repeats the 32-bit value. bandit uses a seemingly random number for the next 32-bit value, but we'll make it work like grackle.
PCCard is used by PowerBook G3 Wallstreet in Open Firmware 2.0.1.
CardBus is probed in New World Macs starting from at least Open Firmware 4.1.9f1 sometime after Open Firmware 3.1.1.
- Create PCIBase from common stuff in PCIDevice.
- Add PCIBridgeBase. These have a primary bus number, secondary bus number, and subordinate bus number which are used to determine if PCI type 1 config cycle should be passed.
- Change PCIBridge to use PCIBridgeBase instead of PCIDevice.
- Add PCICardBusBridge which uses PCIBridgeBase.
For example, Old World Macs have versions of Open Firmware that don't support 512 MB BARs correctly. They may attempt to set such a BAR to 0x90000000 (a 256 MB boundary) instead of 0xA0000000 (the next available 512 MB boundary).
So that you don't see identical messages "Could not restore NVRAM content from the given file."
Instead, one will mention nvram.bin and the other will mention pram.bin.
AdbKeyboard would copy the event into its own fields and set the
changed field, so that we could return the event when register was 0.
However, if a subsequent event was received before ADB polling, the
previous event would be overwritten and lost.
Fix this by maintaining a queue of events, so that we can return
everything since the last poll.
We were using an empty value on the second byte of the ADB keyboard
register 0, but that maps to the "a" key. This manifested itself
as the Key Caps DA never showing the "a" key as being down.
Switch to a non-existent key for the second byte.
Allows different implementations for different platforms (the JS
build relies on browser APIs to stream disk images over the network).
Setting aside the JS build, this also reduces some code duplication.
Besides generating KeyboardEvents in the SDL event handler and
returning the key state in the register 0 reads of the AdbKeyboard
device, we also needed to generalize the ADB bus polling a bit. We now
check all devices that have the service request bit set, instead of
hardcoding the mouse.
The SDL key event -> ADB raw key code mapping is based on BasiliskII/
SheepShaver's, but cleaned up a bit.
Result of running IWYU (https://include-what-you-use.org/) and
applying most of the suggestions about unncessary includes and
forward declarations.
Was motivated by observing that <thread> was being included in
ppcopcodes.cpp even though it was unused (found while researching
the use of threads), but seems generally good to help with build
times and correctness.
The disk cache is unchanged. data_ptr continues to be only used for the user data sector area for each block. The other sector areas (synch, header, etc.) are filled in while reading.
has_data and get_data exist as a way to bypass data_ptr for parts of the transfer outside the user data sector area of each block. The default behaviour is defined in atabasedevice and is overridden by atapicdrom for the Read CD command. atapicdrom has a flag doing_sector_areas to control the behavior of the get_data method. When the flag is true, the sector_areas, current_block, and current_block_byte are used for selecting the correct data from one of the sector areas. The Read CD command initializes those variables. xfer_cnt remains the total number of bytes to be transferred and is now not necessarily the same as the number of disk image blocks read into the disk cache.
lba_to_msf is used to fill in the header. The values was not verified using a real CD.
Mac OS X just cares about the Mode in the header. For now, only the synch and header and user data areas are filled in. The other areas read as all zeros.
It was possible to corrupt New World nvram using dingusppc setenv command.
- setenv must call get_config_vars to set data_length so that it can calculate free space.
- data_length represents the number of bytes taken by nvram variables including the terminating null for the value (name, '=', value, '\0'). Previously, it did not include the terminating null.
- The list of variables and values ends at a '\0' or at pos == 4096 bytes. Previously, data_length wouldn't get set if pos >= 4096.
- Allow setenv to create new nvram variables.
- Since data_length now represents the total number of used bytes, free_space now represents the actual free space, so use > free_space (instead of >= free_space) to determine insufficient space.
- While parsing nvram variable name, do not read beyond 4096 bytes.
- Use a different error message for each problem that can occur while parsing nvram variable name.
If a nvram variable has CRLF or CR, replace them with LF so each line appears on a new line in the console output.
Also, add indent to each line so that each line appears only in the value column and not in the name column.
- Added an overloaded version of attach_pci_device which takes a suffix string that can be used to make a device name unique so that multiple devices of the same class can be added to a machine. The method returns a PCIDevice which can be easily used to attach more PCI devices if it is a PCIHost.
PCIDevice
- supports_io_space method now uses a flag has_io_space which is automatically set for PCI bridges or PCI devices that have an I/O BAR.
atirage
- Devices that have I/O BARs don't need a supports_io_space method.
mpc106
- Devices that don't have I/O methods don't need a supports_io_space method.
A multi-function device may have functions of the same class (e.g. a USB device with two OHCI functions) but each function should be initialized with a different interrupt pin.
Add PCI bridge and multi-function device support.
Overview:
- A multi-function device is two or more PCIDevices with the same device number but one device is function zero (as with currently implemented PCIDevices) and the other functions have function numbers between 1 and 7. The device number and function number are properties of the PCIDevice's parent PCIHost connection.
- A PCIBridge is a PCIHost (it can connect child PCI devices) and a PCIDevice (it has config space, BARs, and expansion ROM).
- A PCIDevice has Type 0 header. It has 6 BARs.
- A PCIBridge has Type 1 header. It has 2 BARs. The config space registers beginning from offset 0x18 differ from those of a PCIDevice.
Possible future modifications:
- Add a PCICardBus class. It is a PCIHost. It has Type 2 header. It has one BAR. The first 20 bytes match Type 0 and Type 1 headers. These exist in New World Macs. They allow hot-plug of PCI devices.
- Split base PCI registers (first 16 bytes) into a PCIBase class. Type 1 and 2 have two or one BAR but I think all 6 BARs belong in PCIBase class anyway.
- Split PCIHost into two classes: Currently existing PCIHosts (Bandit, Grackle) are PCIHost and PCIRoot (they have the broadcast I/O requests functionality) while PCIBridge is PCIHost only - it can propagate I/O requests but does not originate the broadcast.
- pci_register_mmio_region should maybe return a pointer to a region struct so that it can be used for unregistering or modifying the region's range. This may be useful for PCI bridges which have ranges that may constrain memory BARs of their downstream devices.
PCIDevice
- Moved expansion ROM BAR handling to a separate function pci_wr_exp_rom_bar so that it can be used by both PCI devices and PCI bridges which have the ROM BAR in different locations. It now supports unmapping expansion ROM. Also made exp_rom_bar not writable if there's no ROM.
- Added num_bars field which specifies the number of valid BARs since Type 0, 1, and 2 headers have different number of BARs.
- map_exp_rom_mem now properly unmaps expansion ROM (using new function unmap_exp_rom_mem) before mapping it again.
- Added function set_multi_function which modifies hdr_type to indicate if a device has other functions. This is to be applied only to devices with function number 0.
PCIHost
- When attaching a PCI device, it will check if it's a multi-function device (there exists an attached function that is not zero) and adjust hdr_type of function 0 of the device accordingly.
- Attached PCI bridges are added to a list of PCI bridges attached to the host.
- Added pci_io_read_loop and pci_io_write_loop which loop through attached PCI devices to find one that will perform the action for the given I/O address without logging an error (since some other device might perform the action).
- Added pci_io_read_broadcast and pci_io_write_broadcast which are used by a PCI root (bandit/grackle). They will log an error if the action is not performed. They should probably do a machine check exception to match real Power Macs.
- pci_find_device (used by PCI root) will recursively find a PCIDevice for type 1 config register accesses.
- Logging from PCIHost now includes the name of the PCIHost instead of just "PCIHost" because there can be multiple PCI hosts.
PCIBridge
- Sets num_bars to 2 and hdr_type to 1.
- I/O ranges set in the config registers are handled correctly by pci_io_read and pci_io_write.
- Memory ranges set in the config registers do not currently affect memory mmio regions. It is assumed that Open Firmware and the OS will set the ranges and BARs correctly to allow all BARs to be accessed fully.
bandit, mpc106
- Bandit and Grackle now call pci_io_read_broadcast and pci_io_read_broadcast to pass I/O accesses to downstream PCI devices.
- Chaos is modified to work like Bandit even though it will never have PCI bridges attached or devices that support I/O accesses. It's simpler this way.
PCIHost
- PCIHosts (bandit and grackle) now use device number and function number for identifying attached PCIDevices. A macro DEV_FUN is added to calculate this new slot_id. Bandit no longer uses IDSEL. Grackle no longer uses only device number.
machinecatalyst, machinegossamer, machinetnt
- Use DEV_FUN to attach PCI devices by device number and function number.
While dingusppc only emulates 32-bit Macs (for now), it is possible for a 32-bit Power Mac to use a PCIe card that has 64-bit BARs.
finish_config_bars is added to scan the cfg values of the BARs and determine their type. The type is stored separately so that it does not need to be determined again.
The type can be I/O (16 or 32 bit) or Mem (20 or 32 or 64 bit). A 64 bit bar is two BARs, the second contains the most significant 32 bits.
set_bar_value uses the stored type instead of trying to determine the type itself. It is always called even when the firmware is doing sizing. For sizing, It does the job of setting the bar value so do_bar_sizing is now just a stub.
Every PCIDevice that has a BAR needs to call finish_config_bars after setting up the cfg values just as they need to setup the cfg values. Since they need to do both, maybe the cfg values should be arguments of finish_config_bars, then finish_config_bars() should be renamed config_bars().
- Macros need parenthesis to enforce operation order when expanded.
- Fix device and register numbers in log messages.
- Unmapped I/O space reads don't necessarily return 0xffffffff. That's only for config space reads. Just return 0. Unhandled I/O space read should probably cause a memory check (TEA - Transfer Error Acknowledge) exception as it does on a Power Mac 8600.
dingusppc could not read bytes from offset 1,2,3 or words from offset 2.
dingusppc did not read words from offset 1,3 and longs from offset 1,2,3 in the same way as a real Power Mac 8600 or B&W G3.
This commit fixes those issues.
- Added pci_cfg_rev_read. It takes a 32 bit value from offset 0 and returns a value of the specified size using bytes starting from the specified offset. Offsets 4,5, & 6 wrap around to 0,1, & 2 respectively. The result bytes are in flipped order as required by the read method (so a value of 0x12345678 is returned as 0x78563412)
A real Power Mac 8600 might return a random byte for offset 4, 5, 6 for vci0 but usually not for pci1. A B&W G3 seems to always wrap around correctly. We won't read random bytes, and we won't read a default such as 00 or FF. We'll do the wrap around which makes the most sense because writing 0x12345678 to any offset and reading from the same offset should produce the value that was written.
- Added pci_cfg_rev_write. It takes a 32 bit value from offset 0, and modifies a specified number of bytes starting at a specified offset with the offset wrapping around to 0 if it exceeds 3. The modified bytes take their new values from the flipped bytes passed to pci_cfg_write. When size is 4, the original value is not used since all bytes will be modified.
Basically, those two functions handle all the sizes and all the offsets and replace calls to BYTESWAP_32, read_mem or read_mem_rev, and write_mem or write_mem_rev.
read_mem_rev, as it was used by pcidevice and some other places, could read beyond offset 3 if it were ever passed a reg_offs value that did not have offset as 0. Since the offset was always zero, it would always read the wrong byte or word if they were not at offset 0. Same for read_mem as used by mpc106.
write_mem_rev, as it was used by pcidevice and some other places, could write beyond offset 3 if it were ever passed a reg_offs value that did not have offset as 0. Since the offset was always zero, it would always write the wrong byte or word if they were not at offset 0. Same for write_mem as used by mpc106.
pcidevice:
- The logging macros should be used to handle all config register access logging.
- Unaligned PCI config register accesses will be output as ERROR instead of WARNING.
- The logging macros include the offset and size. They also include the value for named registers or for writes.
- Added MMIODevice read and write methods so that PCIDevice is not abstract if a PCIDevice doesn't override the read and write method since some PCIDevices don't have MMIO.
pcihost:
- Added pci_find_device stub for handling PCI bridges in future commit.
bandit and mpc106:
- PCI host controllers will handle all PCI config access alignment and sizing. A PCIDevice will always access config registers as 32 bits on a 4 byte boundary. The AccessDetails passed to a PCIDevice config read or write method is there only for logging purposes.
bandit:
- Common MMIO code is moved to new BanditHost class so both Bandit and Chaos can use it. PCI related code is moved to new BanditPCI class.
- Simplify IDSEL to/from PCI device number conversion by removing the shift or subtract.
- Remove BANDIT_ID_SEL check. The IDSEL conversion to PCI device number can find the bandit PCI device.
- For logging, make best guess of PCI device number from invalid IDSEL - the result is always reasonable for device 0x00 to 0x0A when accessing config register 0x00 (as one would do when scanning for PCI devices like lspci does).
mpc106:
- Common config space code is put in cfg_setup. It handles extracting the offset.
- Added code to log access to unimplemented config registers of grackle.
- Don't call setup_ram when writing to config registers that setup_ram doesn't use.
- pci_cfg_read calls READ_DWORD_LE_A and pci_cfg_write calls WRITE_DWORD_LE_A. When reading or writing memory that is organized as little endian dwords, such as my_pci_cfg_hdr of mpc106, the function should explicitly state that it's little endian so that the emulator may be ported one day to a CPU architecture that is not little endian.
atirage:
- The changes correctly place user_cfg at byte 0x40 instead of 0x43 and writes the correct byte depending on size and offset.
Use explicit cast when converting large integer types to smaller integer types when it is known that the most significant bytes are not required.
For pcidevice, check the ROM file size before casting to int. We'll allow expansion ROM sizes up to 4MB but usually they are 64K, sometimes 128K, rarely 256K.
for machinefactory, change the type to size_t so that it can correctly get the size of files that are larger than 4GB; it already checks the file size is 4MB before we need to cast to uint32_t.
For floppyimg, check the image size before casting to int. For raw images, only allow files up to 2MB. For DiskCopy42 images, it already checks the file size, so do the cast after that.
First, remove name override for subclasses of HWComponent (Chaos and ScsiBus) because HWComponent has its own name field.
HWComponent name should be set as early as possible in the constructor so it can be used in log messages.
PCIDevice should set name of HWComponent (through MMIODevice) in its constructor, using the name that is given to its constructor.
For Bandit and Grackle, they don't need to set the HWComponent name since its PCIDevice constructor will now do it.
Chaos is not a PCIDevice so it should set the MMIODevice name itself.
Why does PCIDevice have a name that is separate from the HWComponent name?
Usually bit 1 of I/O BARs is 0 since it is reserved, but control has an I/O BAR where this bit is set.
bar_cfg is used to determine the default values of the least significant bits (2 bits for I/O BARs and 4 bits for Memory BARs).
The upper bits of bar_cfg determine which bits can be set and also determines the size of the BAR.
