Short: Low Level Debugger Author: thomas.richter@alumni.tu-berlin.de (Thomas Richter) Uploader: thomas richter alumni tu-berlin de (Thomas Richter) Type: dev/debug Version: 1.107 Requires: An Amiga or even better two. The OS version doesn't matter. Architecture: m68k-amigaos ***************************************************************************** ** ** ** ___________ ** ** */ ** ** /----------- ####### ####### ######## ** ** | ^ | ## ## ## ## ## ## ** ** | _ ## ## ## ## ## Version 1.107 ** ** | | ## ## ## ######## ** ** | --| ## ## ## ## ** ** +-----/ ## ## ## ## ## ** ** | | ####### ####### ## (c) 1993-2023 THOR-Software ** ** ** ** ** ** A Low Level Debugger for all Amiga Systems ** ** ** ***************************************************************************** by _______ _____ _____ __ _ / / / / / / / /_/ //_/_ / /____/ / / /____/ ___ / ____// / _ __ / / / / / / / /|/ / / / /_/ / /_/_/ / / / /____/ / / _/__/_/ /__/|_/_/_/ /_/_/ /_ / _____________________________________________________________________________ About COP COP is a low level debugger for all amiga systems. "Low Level" means it directly plays with the hardware, does not have GUI, but is unbeatable stable. Its main purpose is to debug programs not debugable with the usual "source level" debugger, like hardware device drivers, DOS handlers in action and much more. It supports debugging by the serial port, thru a standard tty terminal (also a COP type 5 terminal is better. A terminal program for amigas is included), and a build-in terminal emulation in case no second computer is available. In case you want to use COP, make sure: -that you really want it. COP is not easy to use! -that you know assembly language. It won't work without! -that you have a lot of good will! EMail: thomas.richter@alumni.tu-berlin.de By the way: You should read the documentation first (before trying to get COP running)! This is even more important if you own a '060 or '040 system. Special hints can be found at the end of the COP.doc file. _____________________________________________________________________________ Changes made in 1.107 - Added support for SegTracker 47.1, that is, COP now also prints symbol and source code/line number information on stack tracebacks and with the SEGM command if available. _____________________________________________________________________________ Changes made in 1.106 - The ftrapcc instructions were not disassembled correctly if they included a data field. - If segment information was printed during a stack traceback, the lines including the information did not accumulate for the pager. _____________________________________________________________________________ Changes made in 1.105 - The fssqrt and fdsqrt instructions were not disassembled correctly, fixed. - Added NOFPUEMU command line option and the ENVI.S option to bypass LineF emulator traps that could be handled by software based FPU emulators such as SoftIEEE. SoftIEEE users should instruct COP to bypass such exceptions to SoftIEEE with the NOFPUEMU command line option. _____________________________________________________________________________ Changes made in 1.104 - The disassembly of fdbcc, pdbcc, cpdbcc (co-processor loop instructions) was not correct and expected a 32-bit displacement where a 16-bit displacement would apply. - The disassembly of the k-factor of fmovem packed decimal to memory was wrong. The data register was not disassembled correctly, and an immediate negative k-factor was not shown correctly. - The disassembly showed a k-factor when reading from packed-decimal, but it has no meaning there. _____________________________________________________________________________ Changes made in 1.103 - The division operator could have reported an overflow even though there was nothing to report. - Searching with memory management disabled (ENVI.M command) could not be aborted. _____________________________________________________________________________ Changes made in 1.102 - In case more than one RTG board was installed, COP did not manipulate the display switches on the boards correctly to get its screen frontmost. - Boards with a software-driven flicker fixer such as the Picasso IV can still not be disabled to show the COP screen since activating the on-board flicker fixer cannot be done without operating system support, but COP should now at least properly ignore such RTG boards. In such cases, please debug over a null-modem cable (which is recommended anyhow). _____________________________________________________________________________ Changes made in 1.101 - The multiplication operator * did not correctly handled signed input and returned nonsense if one of the two factors were negative. _____________________________________________________________________________ Changes made in 1.100 - The FILL command no longer fills across sections and aborts on section boundaries. In particular, it no longer wraps around, and warns if its start position is not in a section - unless memory management is turned off. This idiocracy of former FILL commands could have had desasterous side effects. _____________________________________________________________________________ Changes made in 1.99 - The 68060 processor cannot restore the state when entering the unimplemented FPU exception vector with a NULL state. In such a case, COP restored the NULL state, which trashed the FPIAR (and other status registers). Now COP restores the status registers, though leaves the FPU in the IDLE state. This is neither perfect, but allows to step through FPU exception handlers. _____________________________________________________________________________ Changes made in 1.98 - The disassembly of fmovem of multiple control registers using the immediate addressing mode as source was not correct - fixed. _____________________________________________________________________________ Changes made in 1.97 - COP now also prints DOS volumes, devices and assigns with the new "LIST.V" command. - The NODE command was extended to print Dos devices briefly with NODE.V, or verbosely, with NODE.D. _____________________________________________________________________________ Changes made in 1.96 - The memory setup marked the zero page (the first 4K or 1K) as part of the available memory, potentially causing hits (or COP capturing itself) when a FIND reaches this area. - FIND did not respect the "memory management disable" switch "ENVI.M" and hence could not scan non-memory parts of the system map. _____________________________________________________________________________ Changes made in 1.95 - The P96 monitor switch function was a bit uncareful and crashed the machine with the "Native" video driver loaded. This should be fixed now. _____________________________________________________________________________ Changes made in 1.94 - The previous fix wasn't one.... here is the second attempt for getting the bit instructions disassembled correctly. _____________________________________________________________________________ Changes made in 1.93: - The disassembler incorrectly identified the btst opcodes with PC-relative addressing as "illegal". This has been fixed. _____________________________________________________________________________ Changes made in 1.92: - The LIST and NODE commands now also now resident modules and the resident structure. In particular, LIST.E lists the installed resident modules, both in ROM and those added by the KickTags, and LIST.F lists the kicktags installed in exec. - Memory management was less than ideal if fast memory was fragmented. COP 1.91 and before was only able to identify one big block of FAST_MEM. This release can now handle up to four disjoined blocks. _____________________________________________________________________________ Changes made in 1.91: - When stepping through programs with the "N" command, the debugger now automatically reverts to tracing for instructions that return from subroutines or exceptions (RTS,RTD,RTR and RTE). Note that for branches, "N" still sets a breakpoint because they might be part of a loop. _____________________________________________________________________________ Changes made in 1.90: - This release fixes a race condition in the "region clear" VT102 terminal handler that could have caused an infinite loop when debugging over a serial connection. _____________________________________________________________________________ Changes made in 1.89: - COPs internal crash handling had a bug and failed to restore the interrupt settings when recovering from a COP-crash, leaving the system unusable. - COPs memory testing did not check properly for AbsExecBase accesses and crashed (correctly) on byte- or word-accesses. - COP supports now standard VT-102 terminals (yeah!) as external serial terminals. It auto-detects the terminal-type, so type-3 terminals continue to work. _____________________________________________________________________________ Changes made in 1.87: - MMU handling is now much better. COP tries now to be very conservative when reading MMU descriptors and pushes them out of cache as soon as possible. Further, COP checks now for the availibility of pages from user and supervisor page before attempting to access them. - Exception handling was improved substancially. The new "env.h" command can now be used to re-direct high-memory access errors to the default exception handling, typically the mmu.library. - COP provides now a "high-speed" serial transfer mode using a baud rate of 115200. This rate requires a "type-5" terminal, for example the terminal program within this archive; simple TTYs operate still at 9600. COP is able to auto-detect the terminal baud rate, just set the terminal to 115200 baud and let it go. _____________________________________________________________________________ Changes made in 1.86: - Fixed the "capture" command for the 060 in the early bootstrap phase. COP used its own internal version of the "AttnFlags" to determine the format of the exec stack frame, but what is important here is what exec "believes" about the processor and not what "COP knows". - Renamed QUAT to QUAD and UQUAT to UQUAD (typo fix). _____________________________________________________________________________ Changes made in 1.85: - Fixed several 060 related bugs: - an unnecessary "fnop" trashed the fpiar register and hence made it impossible to debug floating point exception handlers. - the bootstrap code of various boards let exec identify the 68060 as 68010 and hence trashed the CPU check of the resident code of COP. COP runs now the CPU test itself. - since the 68060.library has to replace AddTask(), COPs AddTask() patch was overwritten in case it was made resident before. "RestoreVBR" will now re-install the patch to make sure the proper exception handler is called on a crash. This is a bit touchy and requires the creation of a temporary task. _____________________________________________________________________________ Changes made in 1.83: - Fixed the fmovem  disassembly. - Fixed incorrect FPU exception handling for the 040 and the 060. - Fixed a spurious Bus error which occured on some 040 systems at times the MMU wasn't active (Early bootup) - Fixed many bugs in the "Terminal" program: - Hangs on menu pop-open are history, I hope. (Semaphore deadlocks) - ATR transfer became more stable (but still leaves a lot to be desired) - ATR "autodetection" mode has been added. _____________________________________________________________________________ Changes made in 1.82: - Fixed the (d.32,PC) EA disassembly which was wrong by two bytes. - Added the comment character ";" to the command interpreter. _____________________________________________________________________________ Changes made in 1.81: - Added some workarounds in the terminal driver against a 68030 firmware bug. - Fixed the "find.i" command which was broken. - Fixed printer support. In case of printer trouble, it did not disable printing correctly. - The long mulitplication and division opcodes are now correctly disassembled as "extended". - Fixed a bug in the setup logic which caused the program to report that interrupts have been overwritten on plain 68000 systems. _____________________________________________________________________________ Changes made in 1.80: - COP disables now the caches if the MMU is disabled. This helps preventing some cache related problems on boards with Z-II memory which might not allow burst accesses. - COP autodetects now a 68060 CPU even if the correspoding 060 flag in ExecBase->AttnFlags is not set correctly. The SET060 option is no longer required to correct this manually. - Fixed a bug in the disassembler, did not disassemble 64 bit arithmetics correctly. - Updated the MuForce compatibility hints slightly. Mainly, specify NOGURUPATCH and avoid CAPTURESUPER. _____________________________________________________________________________ Changes made in 1.79: - Fixed the terminal program a lot. The I/O mechanism was very unstable. - Updated the ATerm download protocol. - Partially fixed '060 support of the 1.77 release. This might be still broken, but there's currently nothing I can do about it. - Rewrote the vector setup logic and the autovector monitor. - Updated the disassembler: tst.l ax was disassembled as illegal instead of extended. - Disassembler is now really MMU aware, removed an old cludge. _____________________________________________________________________________ Changes made in 1.77: - Rewrote the IRQ table setup for MC68010 or better. It used to be a mess. - Fixed the cache control commands for the 68040 and 68060. - Added SegTracker support for stack traceback. - Fixed memory manager MMU support. It used to use the SRP instead of the URP for checking table accesses and therefore caused problems with MuGuardianAngel. This release seems still to have problems on some 030 machines if it is installed with SetCPU FASTROM active. I don't know why as I can't reproduce it on my 030 A2000. MuFastROM is recommended anyways. _____________________________________________________________________________ Changes made in 1.76: - Internal release, unpublished. _____________________________________________________________________________ Changes made in 1.75: - Modified the exception handler routines a bit, they no longer block interrupts completely. - Added the DARKEN command line option to blank the screen. - Fixed the memory list scan that failed with remapped low-memory areas. _____________________________________________________________________________ Changes made in 1.73: - The "restorevbr" option did not work correctly and could cause problems on 040 and 060 based machines. - The MMU could have been disabled on startup, forgot to saveback a register. - 030 bus error processing can now handle MuForce instruction emulation in the zero-page. _____________________________________________________________________________ Changes made in 1.72: - Fixed a stupid bug in the 060 support code. COP did not test correctly whether a memory address is really available due to a broken MMU table scanner. _____________________________________________________________________________ Changes made in 1.71: - The 1.69 version disabled the MMU on 030 systems completely. Fixed. - COP can now be made resident even on systems without autoconfig fast memory. - The COP initializer was a bit too picky about the FPU bits and therefore refused to keep resident on 040 and 060 systems. - Fixed 040 FPU related bugs. - Partial RTG support implemented, yeah! _____________________________________________________________________________ Changes made in 1.69: - Fixed the disassembly of some 040 specific instructions, cpushp and others. - Fixed the startup code. - Fixed the documentation, COP was never Enforcer compatible when run on a 040 and 060, unfortunately. It is compatible to "MuForce", which does the same. - The bus error handler tries now to complete missing writebacks of the 040 processor. _____________________________________________________________________________ Changes made in 1.68: - Added 040 and 060 FPU instructions for single and double precision in the disassembler. - Added PLPAR/W instructions. - Overflow handling in the expression evaluation was buggy. - MMU table lookup for 68060 was buggy, worked only for complete 4K tables. - MMU management included, the MMU registers are now written back. (Let's hope this works for the 040 and 060, I haven't tested this) - The MMU URP/SRP registers can now be set by the user. However, be careful with it, this may easely trash the system. - FPU emulator traps haven't been filtered out for the 040. _____________________________________________________________________________ Changes made in 1.67: - Due to a silly mistake, the Enforcer was disabled completely for the 68030 processor.... Argh! _____________________________________________________________________________ Changes made in 1.66: - Fixed a horror bug in the MMU control section. After disabling the MMU, I forgot to flush the ATC and the cache of the 68030. Since the cache of the '020 and the '030 is BEHIND the MMU, it could still contain addresses marked as INVALID and might create an enforcer hit on some systems. This happens most likely in systems with the supervisor stack in chip ram. Fixed. - General cleanup of the vector base register installation routine. - Included the SSPToFast program which *should* be run in the startup sequence for systems without autoconfig fast ram. Moving the SSP to fast ram will increase the performance of the system. This release is mainly to due to Jörg Riemer. Thanks for letting me debug COP on his A1200, and thanks for the hospitality. Changes made in 1.65: - Fixed a bug in the 68851 detection routine. This thing is now getting "metamagical", too. As the rest of the kernal code... Thanks, Dennis, for letting me know. - COP RESTOREVBR restores now, too ,the Alert entry that might have been reset by the Enforcer. - Removed a bug from the symbol hunk parser that didn't respect HUNK_DEBUGs as it should. _____________________________________________________________________________ Changes made in 1.64: - Fixed another set of bugs in the disassembler. The "pmove bad/bac" opcoded did not work. - The disassembler knows now the "fnop" command. - The register saveback routine was faulty for the '020 - it tried to save the non-existing tt0,tt1 registers. _____________________________________________________________________________ Changes made in 1.63: - Fixed a bug in the disassembler. The 0x7100 opcode was not detected as illegal. - Fixed memory addressing bugs in the disassembler. Did not check for validity of addresses. - Fixed an addressing bug in the label manager. Did not check whether an address is valid or not. - COP does no longer run the ROM code for the '060 in case no Enforcer/060 library is installed. - Added support code for the 68851 PMMU coprocessor. Should be detected correctly now. _____________________________________________________________________________ Changes made in 1.62: - Neither the NOMMU nor the NOEMU command line options worked in 1.61. Sigh. - Added MMU and EMU keywords for symmetry. - Fixed just another bug in the '060 support code, thanks to Mark for the hints and the testing. (I guess we're at 4% now). - Added the MMU support for MOVE, FIND and FILL commands. - Improved FIND command, displays the found instructions or bytes. - Made RESTOREVBR a bit smarter. - Updated the included DDT file to contain '060 specific gurus. Thanks, Mark! ____________________________________________________________________________ Changes made in 1.61: - Fixed a lot of bugs in the '060 support code. Thus, the chance that this release operates properly on a '060 has about doubled (from 1% to 2%, to be precise...). - Added NOEMU command line switch to bypass emulator traps. - Added proper support for the '060 '040 MMU and cache. - Debugged the AbsExecBase move.l emulator for the '060. Might actually work by now, but is untested. - Changed some details in MMU handling. - Changed minor details for COP capturing its own bugs. - Added another function to the label manager, showing the label given an address - LABL.F command. - Fixed a bug in the memory manager that did not handle non-writable memory in the correct way. - Fixed a bug in the set-breakpoint logic. - Enhanced the stack traceback with the new label manager. It tries to find out the label names. _____________________________________________________________________________ Changes made in 1.60: - Removed bugs in the breakpoint logic. - Added SegTracker support for the REGS.G and (new) SEGM commands. - Updated MMU/Enforcer support since I finally bought a real 68030. - Added the NOMMU command line switch to disable the MMU bypass. - Added the ability to pre-compile the DDT files for faster parsing with the "PRE" command line option. - Updated the Terminal program a bit. _____________________________________________________________________________ Changes made in 1.59: - Removed a bug in the expression evaluator that made the indirection operator [] unuseable. - Removed a bug in the conditional break points. - Rewrote parts of the system segment. - Added support for the CyberGuard. That's not yet tested, however! - Added more '040 instructions I forgot. - Fixed a bug in the disassembler. - The previous fix of the "NODE" command broke it even more, fixed! - Fixed plain MC68000 support that was broken since 1.56. _____________________________________________________________________________ Changes made in 1.58: -Added two '040 instructions I wasn't aware of. -Rewrote the complete math support, an FPU is no longer required. -Enhanced the "NODE" command, it displays now all additional information it can hold of. -Removed a stupid bug from the "NODE" command. -Added the "LABS" command to view the defined labels. Study the doc file for details. _____________________________________________________________________________ Changes made in 1.57: -Added PAL/NTSC switch. -Added a check for many command that won't work in a splitted display. _____________________________________________________________________________ Changes made in 1.56: -Removed a bug in the printer hardware driver that caused one additional character to be printed. -Added command to eject paper. -Added Enforcer support for the MC68020 and MC68030 CPUs. -Added partial MMU support, however no MMU related commands are available. Enforcer must be run for this purpose. -Added IRQ CopyBack mode. -Fixed several final bugs in the 1.55 release, esp. some MC68060 related, although the support isn't complete yet ('60 FPU support is still somewhat broken, as the FPU part must be re-written anyways). Thomas, March 2023