Brett Gordon [Mon, 18 Dec 2017 13:58:47 +0000 (08:58 -0500)]
coco3: take cmd line for default video mode
Alan Cox [Fri, 15 Dec 2017 13:05:31 +0000 (13:05 +0000)]
65c816-big: We can't build it all -Oi, but we can do the core code
Alan Cox [Fri, 15 Dec 2017 00:09:49 +0000 (00:09 +0000)]
v65c816-big: A development configuration for a 65C816 with loads of RAM
8MB RAM testing work.
Lots to do here (plus some of the core code like the scheduler doesn't really
scale to 128 processes!)
Alan Cox [Fri, 15 Dec 2017 00:09:21 +0000 (00:09 +0000)]
level2: fix build errors from previous changes
Alan Cox [Fri, 15 Dec 2017 00:08:50 +0000 (00:08 +0000)]
v65c816: README
Signals (from kernel side at least) are done
Alan Cox [Fri, 15 Dec 2017 00:08:36 +0000 (00:08 +0000)]
65c816: fix bank != 0 case
Alan Cox [Fri, 15 Dec 2017 00:08:19 +0000 (00:08 +0000)]
65c816: fix lib code when bank != 0
Alan Cox [Thu, 14 Dec 2017 21:55:30 +0000 (21:55 +0000)]
inode: another SDCC size-ism fixup
Alan Cox [Thu, 14 Dec 2017 21:38:31 +0000 (21:38 +0000)]
kernel: another pipe 16bit fix
Alan Cox [Thu, 14 Dec 2017 21:31:10 +0000 (21:31 +0000)]
inode: do some more pipe maths in 16bit
Alan Cox [Thu, 14 Dec 2017 21:25:23 +0000 (21:25 +0000)]
65c816: documentation update
Alan Cox [Thu, 14 Dec 2017 17:10:51 +0000 (17:10 +0000)]
v65c816: turn on more features, fix memory map bugs
Turn on networking core support
Correct the definitions so we are not 0x200 bytes off on the stacks
Move the IRQ DP into $FFxx as it can share with the vectors
Move discard into the upper bits of the stacks where it can be reclaimed
Alan Cox [Thu, 14 Dec 2017 15:22:58 +0000 (15:22 +0000)]
65c816: add hints for process.c and also fix up the waitpid loop
Without this fix we can end up stalling in weird corner cases
Alan Cox [Thu, 14 Dec 2017 15:22:43 +0000 (15:22 +0000)]
65c816: add register hints
Alan Cox [Thu, 14 Dec 2017 15:21:34 +0000 (15:21 +0000)]
65c816: use register ops here and there
We need this to get better performance out of cc65.
We may be able to justify -Oi but the code grows a lot so it may become a
size problem if we did.
Alan Cox [Thu, 14 Dec 2017 14:20:26 +0000 (14:20 +0000)]
lseek: Fix signed/unsigned complaint from cc65
Alan Cox [Thu, 14 Dec 2017 13:54:59 +0000 (13:54 +0000)]
fcntl: Fix F_GETFL bug
Alan Cox [Thu, 14 Dec 2017 13:45:29 +0000 (13:45 +0000)]
kernel: fix typo in process change
Alan Cox [Thu, 14 Dec 2017 13:43:52 +0000 (13:43 +0000)]
6502/65c816: introduce a regptr hint
We need this because cc65 writes *far* better code if you tell it to mark
key structure pointers as register so it can then use zp based offset addressing
on them
Alan Cox [Thu, 14 Dec 2017 13:43:12 +0000 (13:43 +0000)]
z80pack: map is always 8bit.. so use an 8bit value
Alan Cox [Thu, 14 Dec 2017 13:42:49 +0000 (13:42 +0000)]
process: re-arrange code for clarity
Alan Cox [Thu, 14 Dec 2017 13:42:16 +0000 (13:42 +0000)]
kernel: introduce a u_done counter set to 0 in rwsetup
We then use this to clean up psleep_flags_io and some other code. This also
saves us a few bytes
Alan Cox [Thu, 14 Dec 2017 12:58:24 +0000 (12:58 +0000)]
signals: correct SIGCONT, optimize code slightly
Alan Cox [Thu, 14 Dec 2017 12:20:00 +0000 (12:20 +0000)]
filesys: use the getmode helper in i_deref instead
Cleans up a bit, uses 8bit compares and saves us another 26 bytes
Alan Cox [Thu, 14 Dec 2017 12:13:39 +0000 (12:13 +0000)]
i_deref: remove repeated evaluations
Cleans up the code and saves us another 26 bytes as a side effect
Alan Cox [Thu, 14 Dec 2017 12:04:54 +0000 (12:04 +0000)]
kernel: remove a single use variable
Tidying up - and shorter cc65 code. Bizarrely sdcc also can't figure out that
this could just be a temporary so it also saves 12 bytes
Alan Cox [Thu, 14 Dec 2017 11:56:11 +0000 (11:56 +0000)]
syscall_fs: diet lseek another 80 bytes or so
Alan Cox [Thu, 14 Dec 2017 11:48:30 +0000 (11:48 +0000)]
kernel: avoid switch in lseek
Not all our compilers do the smart thing with this, and we get better code using
if statements on those that don't.
Alan Cox [Thu, 14 Dec 2017 11:32:35 +0000 (11:32 +0000)]
process: Fix mask size in ssig
We don't need a 32bit value any more, and using 16bits saves us 100 bytes
(That's how expensive SDCC longs are!)
Alan Cox [Thu, 14 Dec 2017 00:45:14 +0000 (00:45 +0000)]
inodeL small optimization
Pipes always wrap at 18 * 512 bytes so we don't need to do a 32bit comparison
here
Alan Cox [Wed, 13 Dec 2017 23:32:25 +0000 (23:32 +0000)]
kernel: Make signal handling 16bit clean
SDCC makes a complete hash of 32bit maths and the other compilers don't handle
it that well either so rework the signals as 16bit chunks. This saves us a
whopping 350 bytes as well as speeding stuff up.
Also copy over the held bits - something we forgot to do before.
Alan Cox [Wed, 13 Dec 2017 21:31:19 +0000 (21:31 +0000)]
fsck: some cleaning and tuning
Alan Cox [Wed, 13 Dec 2017 20:57:25 +0000 (20:57 +0000)]
65c816: remove some debug
Alan Cox [Wed, 13 Dec 2017 20:36:30 +0000 (20:36 +0000)]
65c816: signal fixes
With this we seem to have working signals
Alan Cox [Wed, 13 Dec 2017 20:18:32 +0000 (20:18 +0000)]
65c816: signal improvements
We now correctly call the vector for the signal and process it. We return to the
correct recovery code, and then fail to recover correctly.
Alan Cox [Wed, 13 Dec 2017 19:16:14 +0000 (19:16 +0000)]
size: add 6502
Alan Cox [Wed, 13 Dec 2017 19:14:50 +0000 (19:14 +0000)]
65c816: handle interrupt race
Handle the 1 instruction window where we interrupt just as we do the RTL to go
back to user space. Without this additional check we see user space as the mode
and instead of executing the RTL after the CLI we end up executing something in
the user bank at that address.
Alan Cox [Wed, 13 Dec 2017 19:13:00 +0000 (19:13 +0000)]
6502: Fix longmp
This is ugly. For the 6502/6509/65C02 and friends we need to just deal with the
8bit S register. For the 65C816 we have to change the low 8bits of S only, while
keeping interrupts off as we might otherwise get swapped out and change S mid
calculation.
Not always my favourite processor for multi-tasking..
Alan Cox [Wed, 13 Dec 2017 14:13:42 +0000 (14:13 +0000)]
z80: crt0 - let binman fill in the header properly
Alan Cox [Wed, 13 Dec 2017 14:12:58 +0000 (14:12 +0000)]
binman: set up sizes correctly
Alan Cox [Mon, 11 Dec 2017 16:02:14 +0000 (16:02 +0000)]
65c816: document tweaks
(Yes this should probably get turned into markdown as well)
Alan Cox [Mon, 11 Dec 2017 15:46:05 +0000 (15:46 +0000)]
65c816: draft porting guide for this architecture
Alan Cox [Mon, 11 Dec 2017 15:45:44 +0000 (15:45 +0000)]
Merge branch 'master' of https://github.com/EtchedPixels/FUZIX
Alan Cox [Mon, 11 Dec 2017 15:45:24 +0000 (15:45 +0000)]
killall: don't kill empty ptab slots!
EtchedPixels [Sun, 10 Dec 2017 14:38:52 +0000 (14:38 +0000)]
Merge pull request #572 from davidgiven/nc200
NC200 floppy boot support
David Given [Sun, 10 Dec 2017 11:16:50 +0000 (12:16 +0100)]
Add README.
David Given [Sun, 10 Dec 2017 11:10:14 +0000 (12:10 +0100)]
Bodge together a floppy-bootable nc200 platform.
David Given [Sun, 10 Dec 2017 10:47:34 +0000 (11:47 +0100)]
Convert bootblock.s to sdasz80 syntax, and add Makefile rules to compile it.
David Given [Sun, 10 Dec 2017 10:27:54 +0000 (11:27 +0100)]
Rename the boot skeleton file to avoid confusing with the kernel bootblock.
David Given [Sat, 9 Dec 2017 23:21:23 +0000 (00:21 +0100)]
Increase the boot partition size to allow three 16kB banks, not just two.
David Given [Sat, 9 Dec 2017 22:50:17 +0000 (23:50 +0100)]
The boot loader now works (probably).
David Given [Sat, 9 Dec 2017 16:02:36 +0000 (17:02 +0100)]
Add a more-or-less working NC200 bootable file system image and auto.prg.
EtchedPixels [Sat, 9 Dec 2017 13:23:48 +0000 (13:23 +0000)]
Merge pull request #571 from beretta42/netd2
netd: let Fuzix choose local port
EtchedPixels [Sat, 9 Dec 2017 13:23:20 +0000 (13:23 +0000)]
Merge pull request #570 from beretta42/killall
utils: add 'killall' to install script
Brett Gordon [Thu, 7 Dec 2017 19:36:34 +0000 (14:36 -0500)]
netd: let fuzix do the local port binding
Brett Gordon [Sun, 19 Nov 2017 03:19:08 +0000 (22:19 -0500)]
netd: unneccessary packet type check.
Brett Gordon [Thu, 7 Dec 2017 18:05:27 +0000 (13:05 -0500)]
utils: add 'killall' to install script
Alan Cox [Thu, 7 Dec 2017 15:11:52 +0000 (15:11 +0000)]
seq: move initializers to SDCC is happy
Alan Cox [Thu, 7 Dec 2017 15:11:24 +0000 (15:11 +0000)]
65c816: fix documentation
Alan Cox [Thu, 7 Dec 2017 15:10:46 +0000 (15:10 +0000)]
v65c816: fix incorrect define
It's right in the asm but the C one is used in swap which we don't yet do..
Alan Cox [Mon, 4 Dec 2017 13:46:39 +0000 (13:46 +0000)]
v65c816: expose the load base to asm for the doexec fixups
Alan Cox [Mon, 4 Dec 2017 13:45:57 +0000 (13:45 +0000)]
v65: put signal vectors in low if the binary is loaded above the normal base
Alan Cox [Mon, 4 Dec 2017 02:14:25 +0000 (02:14 +0000)]
65c816: signal vector copy
Alan Cox [Mon, 4 Dec 2017 01:58:08 +0000 (01:58 +0000)]
fsck: Fix standalone handling of zero indirect
Alan Cox [Mon, 4 Dec 2017 01:57:39 +0000 (01:57 +0000)]
v65c816: actually tick rate is 100Hz
Alan Cox [Mon, 4 Dec 2017 01:49:17 +0000 (01:49 +0000)]
65c816: multiple fixes - pre-empt support
Alan Cox [Sun, 3 Dec 2017 20:27:38 +0000 (20:27 +0000)]
Merge branch 'master' of https://github.com/EtchedPixels/FUZIX
Alan Cox [Sun, 3 Dec 2017 20:27:15 +0000 (20:27 +0000)]
utils: add a SYS5 killall
Not yet tested
EtchedPixels [Mon, 27 Nov 2017 18:31:45 +0000 (18:31 +0000)]
Merge pull request #569 from beretta42/netd2
Quick fixes for netd
Tormod Volden [Sun, 26 Nov 2017 17:50:26 +0000 (18:50 +0100)]
dragon-nx32: Move fonts to high memory
The lower memory segment was running full.
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 25 Nov 2017 18:26:02 +0000 (19:26 +0100)]
Standalone/fsck: Add -y option
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 25 Nov 2017 18:06:47 +0000 (19:06 +0100)]
build-filesystem-ng: Print exit status from tool, not from "system"
Interpret exit status 1 from fsck as non-fatal.
Also list package filenames if running verbose.
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 25 Nov 2017 00:27:42 +0000 (01:27 +0100)]
util: Add a (long) integer-only seq command
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Brett Gordon [Sat, 25 Nov 2017 23:33:42 +0000 (18:33 -0500)]
netd: remote connect is an asynchronous event, kill udp straight away.
Brett Gordon [Sat, 25 Nov 2017 23:13:12 +0000 (18:13 -0500)]
netd: fix copying of query arg in 'dig'
Brett Gordon [Sat, 25 Nov 2017 23:11:15 +0000 (18:11 -0500)]
netd: shut off gethost's alarm after matching answer arrives
Alan Cox [Fri, 24 Nov 2017 21:46:07 +0000 (21:46 +0000)]
65c816: complete support for signal handling in syscall
Pure IRQ/context switch time handling is not yet debugged
Alan Cox [Fri, 24 Nov 2017 20:34:09 +0000 (20:34 +0000)]
kernel.def: Fix dumb error
Alan Cox [Fri, 24 Nov 2017 20:33:20 +0000 (20:33 +0000)]
65c816: gets the signal handling path via syscall closer
We now call the signal handler correctly, but the handler return blows up
Alan Cox [Fri, 24 Nov 2017 20:33:00 +0000 (20:33 +0000)]
6502: correct signal stubs
Alan Cox [Fri, 24 Nov 2017 17:54:09 +0000 (17:54 +0000)]
65c816: pass the correct A X Y values to the 6502 user space
We don't use these but we do specify them so we should follow what 6502
expects
Alan Cox [Fri, 24 Nov 2017 17:53:19 +0000 (17:53 +0000)]
6502: add more apps
Alan Cox [Thu, 23 Nov 2017 23:20:47 +0000 (23:20 +0000)]
lib: Install driveiwre.h
Alan Cox [Thu, 23 Nov 2017 23:07:43 +0000 (23:07 +0000)]
fsck: Add fixes to standalone fsck from native
Alan Cox [Thu, 23 Nov 2017 23:05:19 +0000 (23:05 +0000)]
fsck: Fix sign when reporting blocks and inodes and some const bits
With this fixed we report valid numbers on a big fsck. There is a weird 6502
oddity that it reports "0" for the memory pool, but I think that's a cc65
bug not an fsck one.
Also fix the path allocating on the recursive search. We now don't buffer
overrun given a long path.
Alan Cox [Thu, 23 Nov 2017 23:04:49 +0000 (23:04 +0000)]
du: 6502 support
Alan Cox [Thu, 23 Nov 2017 22:22:42 +0000 (22:22 +0000)]
6502: correct crt0
The older compiler when we forced fastcall makes main() fastcall. The current
compiler which we can now use without patches makes make() vararg for some
odd reason. That means we need to leave argv on the stack.
With this set the 65c816 userspace actually seems useable
Alan Cox [Thu, 23 Nov 2017 21:45:08 +0000 (21:45 +0000)]
filesys: Fix corruption and buffer leak on bad input address
Alan Cox [Thu, 23 Nov 2017 02:27:32 +0000 (02:27 +0000)]
65c816: further work on things like return paths and fork() corrections
We still have a weird failure on stat() but a lot more stuff seems to
want to behave now
Alan Cox [Thu, 23 Nov 2017 02:27:01 +0000 (02:27 +0000)]
v65c816: ensure the failing trap handler sets a8/i8
Alan Cox [Thu, 23 Nov 2017 02:26:34 +0000 (02:26 +0000)]
v65c816: Fix hd driver to return correct value
Alan Cox [Thu, 23 Nov 2017 02:26:23 +0000 (02:26 +0000)]
v65c816: remove some debug
Alan Cox [Thu, 23 Nov 2017 02:25:56 +0000 (02:25 +0000)]
devio: correctly mark failed read block as unusable
Alan Cox [Thu, 23 Nov 2017 02:25:41 +0000 (02:25 +0000)]
filesys: handle 8bit bools
Alan Cox [Wed, 22 Nov 2017 17:55:32 +0000 (17:55 +0000)]
6502: Add more build Makefiles
"make" now builds a fairly substantial set of tools
Alan Cox [Wed, 22 Nov 2017 17:55:02 +0000 (17:55 +0000)]
as09: build on 6502
Alan Cox [Wed, 22 Nov 2017 17:54:28 +0000 (17:54 +0000)]
6502: Fix quiz.c
Alan Cox [Wed, 22 Nov 2017 17:54:08 +0000 (17:54 +0000)]
6502: Add V7 makefiles
Alan Cox [Wed, 22 Nov 2017 17:53:38 +0000 (17:53 +0000)]
6502: Make clear build on cc65
Alan Cox [Wed, 22 Nov 2017 17:53:07 +0000 (17:53 +0000)]
6502; Build smallC on 6502