2007-07-05 16:02:05 +00:00
|
|
|
TODO
|
|
|
|
----
|
2007-11-27 20:34:05 +00:00
|
|
|
|
2007-11-27 19:44:48 +00:00
|
|
|
[ ] ask jbe about:
|
|
|
|
./commands/Makefile:40:obj-$(CONFIG_SIMPLE_PARSER) += setenv.o
|
|
|
|
[ ] bark on partition not ending on flash sector boundaries
|
2007-07-05 16:02:05 +00:00
|
|
|
[ ] distclean doesn't work without a config
|
2008-04-04 11:05:03 +00:00
|
|
|
[ ] Every driver should have a remove function.
|
|
|
|
(It must have. Currently there is no provision to given to remove my
|
|
|
|
interrupt handlers and to reset all hardware in use, before the target
|
|
|
|
OS is called. csc 21.03.2008 19:50:34)
|
|
|
|
[ ] Global shutdown_drivers() function is needed to carefully clean up all
|
|
|
|
interrupt handlers and reset hardware to initial post reset or otherwise
|
|
|
|
save state.
|
|
|
|
(Ideally the hardware is set back to post-reset state after this call.
|
|
|
|
Run-up and run-down should be reentrant, so if bootm is failing the system
|
|
|
|
can re-initialize devices. csc 21.03.2008 19:48:49)
|
2007-07-05 16:02:12 +00:00
|
|
|
[ ] Clean up make system. Currently I think there are many things from the
|
2009-12-15 08:11:09 +00:00
|
|
|
Linux make system which are not needed for barebox.
|
2008-04-04 11:05:03 +00:00
|
|
|
(Please retain definitions and related for module loading and relocation, so
|
2009-12-15 08:11:09 +00:00
|
|
|
that special barebox modules containing an ELF object can be linked to the
|
|
|
|
running barebox. csc 21.03.2008 19:52:49)
|
2007-07-05 16:02:05 +00:00
|
|
|
[ ] get/set for nonextisting parameters crashes
|
2007-07-05 16:02:12 +00:00
|
|
|
[ ] There is a xmalloc function which panics when out of memory. Use this
|
|
|
|
function where we don't want to check for oom. Same applies to
|
|
|
|
strdup/xstrdup.
|
|
|
|
[ ] cleanup common.h. There are far too many things in it where we better have
|
|
|
|
a seperate header file.
|
2010-07-01 14:59:21 +00:00
|
|
|
(Use doxygen to track header file hierarchy)
|
2007-07-05 16:02:12 +00:00
|
|
|
[ ] Rewrite network stack. The one-loop-for-all-protocols stack is just too
|
|
|
|
ugly and big and uses thousands of global variables. Identify common
|
|
|
|
functions and make a loop per network protocol
|
|
|
|
[-] Cleanup cpu/*. Many functions there are not cpu specific. For example the
|
|
|
|
cache functions for arm are common for most arm processors. (done for ARM)
|
2008-04-04 11:05:03 +00:00
|
|
|
(I will check this for m68k arch - csc 21.03.2008 19:56:24)
|
2007-07-05 16:02:05 +00:00
|
|
|
[ ] Move SoC specific header files from include/ to include/asm/arch/
|
|
|
|
[ ] Several .c/.h files do not have GNU/copyright headers.
|
2007-07-05 16:02:12 +00:00
|
|
|
[ ] The cramfs driver currently uses direct memory accesses instead of read().
|
|
|
|
This disqualifies it for real block devices.
|
2007-07-05 16:02:05 +00:00
|
|
|
[ ] Add documentation for:
|
|
|
|
[ ] driver model
|
|
|
|
[ ] device parameter
|
|
|
|
[ ] big picture
|
2008-04-04 11:05:03 +00:00
|
|
|
[ ] Add example driver and filesystem for reference purposes. Or denote a driver
|
|
|
|
and a filesystem to be the reference.
|
2007-07-05 16:02:08 +00:00
|
|
|
[ ] realloc seems to crash without further notice when out of memory.
|
2007-07-05 16:02:12 +00:00
|
|
|
[ ] implement a mmap() function. Of course we cannot really mmap without mmu,
|
|
|
|
but for memory mapped devices like nor flash we could provide a pointer to
|
|
|
|
it. With this we would not have to copy uimages to memory.
|
2007-11-08 17:23:44 +00:00
|
|
|
[ ] command line editing is somewhat broken
|
2007-11-15 17:43:06 +00:00
|
|
|
[ ] how to select a ethernet channel for tftp when more than one interface
|
|
|
|
of this type exists in the system?
|
2008-04-04 11:05:03 +00:00
|
|
|
[ ] Cleanup headers. I propose to rearrange the files in a way to reflect the
|
|
|
|
organisation of the source tree. This requires changing the include paths
|
|
|
|
in nearly all files, therefore a merge window for this change must be
|
|
|
|
opened and merge in of all other topic branches must happen before.
|
|
|
|
Week 23 or 42 issue? csc, 21.03.2008 19:36:06
|
|
|
|
[ ] Merge similiar information into a single common and a single arch dependant
|
|
|
|
file. E.g. all those general typedefs should be located in include types.h
|
|
|
|
and arch/asm/types.h. csc 21.03.2008 19:38:57
|
|
|
|
[ ] Get rid of the current linux include dir. It should only contain information
|
|
|
|
required to boot linux. csc 21.03.2008 19:39:02
|
|
|
|
[ ] Unify source headers to show up the GPL lic stuff followed by a doxygen file
|
|
|
|
description.
|
|
|
|
[ ] Enhance and complete doxygen comments.
|
|
|
|
[ ] Eliminate all #ifdef CONFIG_??? in C code. Disabled APIs should be defined
|
|
|
|
as empty macros, so that code is removed by the compiler, not the
|
|
|
|
preprocessor.
|
|
|
|
The #ifdefs should be located in their related header files only.
|
|
|
|
[ ] Create an example board, which implements all functions stubs required
|
2009-12-15 08:11:09 +00:00
|
|
|
to compile a full barebox build. Functions are just stubs containing a
|
2008-04-04 11:05:03 +00:00
|
|
|
panic() call - Real code must be filled in accordingly.
|
|
|
|
Can be copied as starting point for a new target board.
|
|
|
|
[ ] Add generic PCI bios code and a PCI bus driver model. The driver should
|
|
|
|
be configured by an appropriate platformconfig. Detected PCI buses and
|
2010-07-01 14:59:21 +00:00
|
|
|
devices should be represented as a hierarchical tree in /dev
|
2008-04-04 11:05:03 +00:00
|
|
|
[ ] Work out, how to mount driver on PCI devices. The mechanism should be
|
|
|
|
transparent to the driver code.
|
2007-11-08 17:23:44 +00:00
|
|
|
|
2007-07-05 16:02:05 +00:00
|
|
|
DONE
|
|
|
|
----
|
|
|
|
|
2009-12-15 08:11:09 +00:00
|
|
|
[x] Rename vmlinux -> barebox
|
2007-07-05 16:02:05 +00:00
|
|
|
[x] Implement current work directory
|
2007-07-05 16:02:08 +00:00
|
|
|
[x] ARCH=linux should catch ctrl-c
|
|
|
|
[x] Implement 'rm'
|
|
|
|
[x] the mount command currently does not accept a full path (i.e. /dev/nor0) but
|
|
|
|
only a device id string (nor0). This sucks.
|
|
|
|
[x] FS support is not optional eventhough the Kconfig system claims it.
|
2009-12-15 08:11:09 +00:00
|
|
|
[x] barebox used to have support for different consoles. The old code was huge and
|
2007-07-05 16:02:08 +00:00
|
|
|
ineffective, so I removed it. Reimplement it using the driver model.
|
|
|
|
[x] Mount without options should show mounted filesystems
|
|
|
|
[x] Add/fix help texts for commands
|
|
|
|
[x] mw is broken. Add common option parsing for mm/md
|
2007-07-05 16:02:12 +00:00
|
|
|
[x] Implement a info command for malloced space. Use it to find memory holes.
|
|
|
|
[x] patch hush.c so that you can get/set device parameters with dev.param= and remove
|
|
|
|
get/set applets
|
2008-04-04 11:05:03 +00:00
|
|
|
[x] rsc: switching on CONFIG_AUTO_COMPLETE fails, because env_complete()
|
|
|
|
isn't there
|
|
|
|
[x] The header files in include/asm-linux/ are i386 specific. Make it work on
|
|
|
|
your i-Book (bigendian...)
|
|
|
|
[x] Globbing support? Might be too expensive, but on the other hand, if you
|
|
|
|
have the space, why not?
|
|
|
|
[x] Implement protect/unprotect support for cfi devices.
|
|
|
|
[x] Implement a command 'countdown' or similar. It should delay the boot
|
|
|
|
process and interrupt it if a certain string is received. The string could
|
|
|
|
be any key, ctrl-c or a certain string. Maybe like this:
|
|
|
|
countdown -m msg -t timeout -x [ctrl-c|anykey|string]
|
|
|
|
If done, remove the corresponding stuff from common/main.c
|
2010-07-22 03:00:13 +00:00
|
|
|
[X] Board support should go to arch/*/boards/*
|