changelog

思科路由器仿真器,用来仿7200系列得,可以在电脑上模拟路由器-Cisco router simulator, used to fake a 7200 series can be simulated

15-Jan-2006
-----------
  - Added PCI I/O space management.
  - Added a fake Cirrus Logic PD6729 PCI-to-PCMCIA host adapter.
  - Cleanups in NPE initialization code.
  - Fixed some bridge problems with the NPE-300.
  - Code factorization in the C7200 initialization module.
  - Generalization of C7200-IO-FE driver for all NPE types.
  - Added Fault History support.
  - Added chassis selection (standard or VXR)
  - Finally found the bug in the non-jit emulation code! It was a cast problem
    in mult/multu instructions.

16-Jan-2006
-----------
  - Added generic hash tables.
  - Added "instruction lookup tables" (ILT), which greatly improve performance
    of non-JIT mode.
  - Added instruction statistics for the non-JIT mode ("Ctrl-] + j")

17-Jan-2006
-----------
  - Added a port for non-x86 hosts. The emulator ran successfully on a
    Tru64 Unix box (Alpha CPU)! It should also be able to compile and run
    on MacOS X (PowerPC).

24-Jan-2006
-----------
  - Added support for DMA channels on Galileo GT64k controllers since it seems
    that they may be used by IOS (NPE-400 for sure, other NPE?).
	
25-Jan-2006
-----------
  - Added Linux Raw Ethernet driver to the NetIO infrastructure.
  - Hmm, I'm probably missing something in the DMA/SRAM interaction and
    endianness... 

26-Jan-2006
-----------
  - The DMA problem is finally fixed, in fact the SRAM can be accessed with
    byte-swapping.
  - Added an unicast limiter to the DEC21140, to receive only trafic directed
    to the virtual machine (+ multicast/broadcast). This is done by analyzing
    the setup frames (MAC address filter).

27-Jan-2006
-----------
  - The Linux Raw Ethernet driver needs promiscuous mode to be enabled to 
    receive appropriate packets, added the required code.
  - Added endianness definitions, and replaced hton[sl]/ntoh[sl] by the new
    definitions (for exchanges between VM and host).

28-Jan-2006
-----------
  - Added parser infrastructure.

29-Jan-2006
-----------
  - Added a virtual bridge system.
  - Release 0.2.3b

31-Jan-2006
-----------
  - Fixed a stupid bug in DMA handling, which in certain conditions can
    cause an infinite loop.

01-Feb-2006
-----------
  - Added an "instruction block hash" which avoids lookups in the instruction
    red-black tree. The IBH table takes only 16 Kb of memory.
  - Added CRC-12 and CRC-16 functions for future use.
  - Added a "gen_eth" NetIO driver, which allow to access ethernet interfaces
    in a portable way (tested on Linux and Cygwin).
  - Fixes to the PowerPC build on Darwin thanks to Olivier Cahagne.
  - Minor cosmetic changes to display supported/unsupported PA and NPE.
  - Release 0.2.3c

06-Feb-2006
-----------
  - Added working Serial interfaces (PA-4T+) !
  - Bug: CDP not working with serial interfaces ("debug cdp events" reports
    a checksum problem).

07-Feb-2006
-----------
  - Started the 0.2.4 release.
  - Modified the command line parser to allow many NIO per PA (for example
    for serials).
  - A lot of code cleanups in the C7200 initialization.
  - Added PA-8T support.
  - To fix: CDP with Serial interfaces: using cp_len (instead of cp_len+4)
    in packet rx fixes the problem but breaks IP. Probably something related
    to padding, see how a CDP packet is sent (special flag in tx ring ?)
  - Added virtual "beqz" instruction.
  - Added the "-X" option, allowing to use host memory directly instead of
    mapping a file to simulate RAM (faster).
  - Beginning of code reorganization to allow multiple instance contexts.

08-Feb-2006
-----------
  - Added a basic virtual Frame-Relay switch. Unfortunately, LMI must be
    implemented to maintain links up.
  - Finally found the problem related to CDP! It was simply a padding problem.

09-Feb-2006
-----------
  - Add support of promiscuous mode for the DEC21140 "unicast limiter":
    promiscuous mode is used in bridging configurations.
  - Fixed another bug with the serial interfaces: it seems that there is
    a "length substractor" for the txring.

10-Feb-2006
-----------
  - Played a little with the PA-POS-OC3, without success.
	
14-Feb-2006
-----------
  - Finally implemented the basic LMI stuff (only ANSI Annex D). Fortunately,
    IOS rocks and uses auto-detection, so no extra configuration is needed.
  - Integrated a Makefile patch from Max Khon for FreeBSD.
  - Updated documentation.
  - Release 0.2.4.

15-Feb-2006
-----------
  - Fixed a cosmetic bug when displaying MAC address at startup.
  - Fixed a stupid bug in "srav" instruction in x86_64 JIT code. The
    arithmetic shift operation was done on a 64-bit basic instead of 32-bit,
    causing bad propagation of the sign-bit.
  - Added support for multiple instances (no CLI present yet).

17-Feb-2006
-----------
  - Fixed 2 bugs in Frame-Relay switch: a crash occured when receiving
    a packet gave an error, and if many DLCI with the same ID on different
    interfaces were used, only one was announced on one interface.

18-Feb-2006
-----------
  - Began work on advanced configuration parsing (instances).

19-Feb-2006
-----------
  - Is the count/compare mechanism pertinent ?

23-Feb-2006
-----------
  - Continued work on advanced config parsing (instances,NIO).
  - Added registry infrastructure.
  - Fixed some mistakes in documentation.

28-Feb-2006
-----------
  - Continued work on advanced config parsing (NIO).
  - Added an IRQ counter.
  - Continued debugging on recurrent crashes problem.
  - Added proper checks to mts32_raw_lookup (at least it avoids the
    coredump...)
  - Added display of all CP0 registers (except TLB).

02-Mar-2006
-----------
  - Hardened a little the IRQ subsystem, but this doesn't fix the problem.
  - Added a debugging mode to track jumps to address 0 (x86_64 only!).
  - Fixed the teq/teqi instructions in amd64/x86 modes that were buggy.

04-Mar-2006
-----------
  - Added some assembly optimized routines for x86 hosts.
  - Fixed a stupid bug in mts32_raw_lookup (op_size and op_type were
    swapped).
	
06-Mar-2006
-----------
  - Added Unix and Null NETIO types for config files.
  - Fixes for incorrect behavior in virtual memory access routines, in
    case of exception.

08-Mar-2006
-----------
  - Removed the configuration file parsing, in favor of a remote control
    system for use with the dynagen project.
	
09-Mar-2006
-----------
  - Added basic hypervisor infrastructure.
  - Added basic NIO code to the hypervisor commands.
	
10-Mar-2006
-----------
  - Added the hypervisor NIO code.
  - Added Frame-Relay and ATM switches to the hypervisor commands.
	
11-Mar-2006
-----------
  - Played (successfully) with OIR (Online Insertion and Removal). OIR
    will require device deletion...

12-Mar-2006
-----------
  - Fixed EEPROM code which used global variables.
  - Added the necessary code for basic OIR.
  - Added PCI device removal.

15-Mar-2006
-----------
  - A lot of work on C3600 integration (although it is not functional).
  - Added a basic ns16552 driver.

17-Mar-2006
-----------
  - Added some hypervisor commands for PA setup and OIR.
	
20-Mar-2006
-----------
  - Hardened the Mueslix driver: after OIR, a "no shut" on a new interface
    crashed the virtual router. Now, IRQ status is managed appropriately.

24-Mar-2006
-----------
  - Added ptask (periodic task) removal.
  - Added a basic (nearly empty) C3600 IO FPGA chip.

25-Mar-2006
-----------
  - Added a NIO RX multiplexer.
  - Added NIO unset operation for PA.

26-Mar-2006
-----------
  - Modified the virtual ATM/Frame-Relay switches and NIO bridge to use
    the new NIO RX multiplexer.
  - Added the NIO bridges to the hypervisor commands.
  - Fixed a stupid bug in hypervisor ATM/FR switch command, the VC were
    badly parsed.
  - Fixed a memory leak in the hypervisor module (tokens not freed).

27-Mar-2006
-----------
  - Added appropriate locks to the ATM and Frame-Relay switches.
  - Added VC removal for ATM and Frame-Relay switches.

28-Mar-2006
-----------
  - Added registry management to the ATM/FR switches and NIO bridge.

30-Mar-2006
-----------
  - Added a refcount system to the NIO RX multiplexer.

31-Mar-2006
-----------
  - Added NIO removal for NIO bridges.
  - Added NIO bridge removal.
  - Modified the NIO bridge code to use the registry infrastructure.
  - Same as above, for ATM and Frame-Relay switches.

03-Apr-2006
-----------
  - Added the required infrastructure for online removal.
  - Added hypervisor commands for PA removal and to display C7200 hardware.
  - Fixed a bug in PCI device removal.
  - It's now possible to change cards on the fly!
  - The Mueslix driver was incorrectly modified: the TX ring scanner must
    run even if the NIO is not defined.
  - Modified the DEC21140 driver to report a Link Down when the NIO is not
    defined.

04-Apr-2006
-----------
  - Added packet dump (for debugging) to NetIO receive/send operations.

08-Apr-2006
-----------
  - The SMP bug seems to be fixed! The irq_cause field is now manipulated
    atomically (inline asm for x86 and x86_64, pthread mutex on other 
    machines).

12-Apr-2006
-----------
  - Added a VM abstraction layer (for different platforms).
  - Fix of X86 assembly file for Cygwin.
  - The IOS config file is now saved with a specific filename from the
    VM instance.

13-Apr-2006
-----------
  - Added per-instance logfile.

14-Apr-2006
-----------
  - Fixed a stupid bug in non-JIT mode with IRQs.

17-Apr-2006
-----------
  - bugfix: "null" NIO creation didn't record the NIO in registry.	
  - bugfix: mueslix logging.
  - bugfix: adding an NIO to the RX listener list (bad double linked-list)
  - bugfix: bad fgets() use in NIO bridge and ATM/FR switches.

18-Apr-2006
-----------
  - bugfix: the same NIO could be recorded many times in the NIO RX
    multiplexer due to bad design in NIO adding.

19-Apr-2006
-----------
  - Beginning of work on AMD Am79c971 ethernet controller (used by PA-4E
    and PA-8E).

20-Apr-2006
-----------
  - The emulation of AMD Am79c971 seems to be working very well.
  - Modified the physical memory dump function to use the VM log file.

21-Apr-2006
-----------
  - Modified the PA-4E/PA-8E EEPROM definitions to work with VXR midplanes.
  - PA-4E is now usable.

23-Apr-2006
-----------
  - Added "shutdown" operation for PA-4E/PA-8E drivers.

24-Apr-2006
-----------
  - Enhanced lock file management (with POSIX locks).
  - The devices are now dependent from VMs instead of being dependent
    of CPU groups. This makes a lot of things cleaner.

27-Apr-2006
-----------
  - Added IRQ clearing with GT64K (DMA transfers).
  - Study of GT64120 controller to understand address decoding process.
	
28-Apr-2006
-----------
  - Fixed a bug with NPE-175/NPE-225 PEM EEPROM selection.
  - The NPE-400 is now able to support 512 Mb of DRAM! (the supplemental
    memory is in fact considered as IO memory)
  - Added a safety check with RAM size.
  - The CPU identifier (PRID register) is now fixed depending on the NPE
    board.

01-May-2006
-----------
  - Modified device subsystem to order devices by physical addresses,
    optimized dev_lookup/dev_lookup_next.
  - Added infrastructure for various object cleanup.
  - Added "shutdown" code for SRAM, NVRAM, Bootflash, Zero, remote control
    devices.
  - Added CPU deletion support (incomplete).

02-May-2006
-----------
  - Added PCI bus removal.
  - Added "shutdown" code for GT64010 and GT64120.
  - Added CFC0/CTC0 instructions (R7000) and handling for associated 
    registers. TODO: customized vector spacing.

04-May-2006
-----------
  - Introduction of the new MTS64 subsystem!
  - Note: incorrect uses of assert(), especially in insn_lookup.c module.

05-May-2006
-----------
  - Added a free list of chunks for MTS64 + some basic optimizations.
  - Added basic PCI/HT stuff for SB-1 processor (dev_sb1_pci.c module).
  - Added AP1011 (Sturgeon HyperTransport-PCI Bridge) device.
  - Fixed masks for MFC0/MTC0 instructions, and added new definition for MFC0
    to access set 1.
  - Fixed incorrect uses of assert().

07-May-2006
-----------
  - Modified the console handling to be more efficient. 
  - Added global invalidation of the MTS64 cache.
	
08-May-2006
-----------
  - Added selective invalidation of the MTS64 cache (for TLB management).
  - Added MTS64 statistics.
  - Modified the MTS64 hash settings. Origin: shift: 12, hash_bits: 16.
    Replaced by shift: 16, hash_bits: 12 -> more efficient and consumes
    less memory! (logical since IOS uses large TLB entries).
  - Modified cp0 module to use an abstraction layer for MTS access, allowing
    32 or 64 bit modes.

09-May-2006
-----------
  - Fixed MTS64 on x86_64 machines.
  - The clock divisor can now be set per VM and with an hypervisor command.

10-May-2006
-----------
  - Added "shutdown" code for RAM and ROM devices.
  - Added VM object dump ("Ctrl-] + o")

11-May-2006
-----------
  - Bugfix: missing initialization of the address length parameter of accept()
    in the main hypervisor module.
  - Bugfix: missing initialization of registry memory pool (noticed with
    Valgrind).
  - Added NPE and midplane selections to the hypervisor.

12-May-2006
-----------
  - Added NIO deletion to the hypervisor.
  - Fixed status message in hypervisor for object deletion.
  - Added "sub" instruction (without exception support).

14-May-2006
-----------
  - Fixed CPU state change. It is now possible to reboot at any time using
    "Ctrl-] + k".
  - Added VTTY deletion.
  - Added "c7200 stop" command to the hypervisor (not finished).
  - The memory-mapped devices of port adapters are now enabled dynamically,
    depending on the PCI BAR (Base Address Registers) settings. This avoids
    use of hardcoded values.
  - Integrated a patch from Philipp Brenner, which fixes ELF loading problem
    on Cygwin when default text file type is set to DOS. Many thanks to him.
    (symptom was: "load_elf_image: elf_begin: I/O error: raw read")

15-May-2006
-----------
  - Added "c7200 set_config" and "c7200 set_mac_addr" hypervisor commands.
  - Fixed a stupid bug in PCI bus removal.
  - Added more complete shutdown code.

16-May-2006
-----------
  - Fixed uninitialized MTS64 allocated entries (seen with Valgrind).
  - Added missing unmapping of memory-mapped files (reported by Greg).
  - The VTTY shutdown doesn't close stdin anymore (causing terminal
    problems at exit).  
  - Fixed the shutdown procedure, added VM_STATUS_SHUTDOWN as status
    for a VM. Adapted the virtual CPU synchronization.