Home | History | Annotate | Line # | Download | only in arc
TODO revision 1.10 
      1 $NetBSD: TODO,v 1.10 2000/05/09 07:30:14 soda Exp $
      2 
      3 To do list (not in any particular order).
      4 
      5     o	XXX sudden hang up in a few minutes or dozens of minutes.
      6 	2000 Mar 17 -current is OK. Mar 29 is NG.
      7 
      8     o	Move the RO and WIRED attribute from the pte to the pv table.
      9 	This saves four instructions in the tlb miss handler.
     10 
     11     o   Boot. Standalone boot program instead of booting the kernel directly?
     12 
     13     o   Find out why bitmap load to S3-928 flashes screen. (X server)
     14 	Know why (enable linear mode). Need S3 info.
     15 
     16     o   Can we have 32 double registers?
     17 
     18     o	64bit kernel/userland
     19 
     20     o	repair DeskStation and Algor support
     21 
     22 	-   64bit device physical address space for Tyne (see OpenBSD pmap.c)
     23 
     24 	-   requires bus_dma
     25 
     26     o	source code structure is quite obsolete,
     27 	general clean up is needed as nisimura-san suggested.
     28 	especially:
     29 
     30  	-   introduce struct platform and remove ugly ``switch (cputype)''
     31 	    in many places.
     32 
     33 	-   redesign interrupt handler framework to be flexible for
     34 	    possible variations; must be sane and useful for R4030/R4230
     35 	    'local' devices and any combinations with ISA/EISA/PCI
     36 
     37 	-   implement and use bus_dma, keeping in mind TLB assisted VDMA
     38 	    that R4030/R4230 chip provides
     39 
     40 	-   clean up bus_space implementation, remove inb/outb
     41 
     42 	-   try to interporate HZ; may be possible for R4030/R4230 system
     43 	    or 'hardclock() by R4000' system.  Clock resolution of 100Hz
     44 	    without any interporation is substandard
     45 
     46     o	overblocking on interrupt handler, and related problems
     47 
     48 	-   SR_INT_IE should be enabled before calling hardclock().
     49 	    Since this is not done currently, spllowersoftclock()
     50 	    on hardclock() doesn't have effect, and softclock() is
     51 	    handled with all interrupt disabled in this case.
     52 		-> overblocking, possibly causes missing hardclock()
     53 
     54 	-   MIPS3_CLKF_BASEPRI() doesn't work correctly,
     55 	    when MIPS_INT_MASK_5 (== MIPS_INT_MASK_CLOCK) is disabled.
     56 		-> micro optimization on hardclock() doesn't work.
     57 		   but currently this may make hardclock() latency better
     58 		   due to above SR_INT_IE problem.
     59 	    s/MIPS_INT_MASK/MIPS3_INT_MASK/ makes this work, although tricky.
     60 
     61 	-   if (ipending & INT_MASK_REAL_DEV) == 0,
     62 	    softnet() and softclock() are handled with all interrupt disabled.
     63 		-> overblocking, possibly causes missing hardclock()
     64 
     65 	-   softclock() is handled with softnet() disabled.
     66 		-> slightly overblocking
     67 
     68 	-   `netisr' handling in netintr() implies potential race condition.
     69 	    `netisr' access should be protected by splnet().
     70 	    currently this is not real problem due to above overblocking.
     71 
     72 	    `ssir' handling on many mips ports has same problem.
     73 	    It should be protected by splnet() or splserial() or splhigh()
     74 	    (depends on the highest interrupt level which sets `ssir').
     75 	    Since `ssir' is accessed not only by setsoftnet() but also
     76 	    by setsoft(), setsoftnet() should protect `ssir' by splserial()
     77 	    or something. (i.e. priority level which setsoft() will be called)
     78 	    Probably, it is better to split `ssir' variable for each
     79 	    priority level.
     80 	    also, _clearsoftintr() should be called before `ssir' access.
     81 	    currently this is not real problem due to above overblocking.
     82 
     83 	-   INT_MASK_REAL_DEV should be removed
     84 
     85     o	it is better to always disable MIPS_INT_MASK_CLOCK.
     86 	those are the points which should be fixed:
     87 	mips_idle: li t0, (MIPS_INT_MASK | MIPS_SR_INT_IE)
     88 	machdep.c: curpcb->pcb_context[11] = MIPS_INT_MASK | MIPS_SR_INT_IE;
     89 	spl0()
     90 	splnone()
     91 
     92 	-   MIPS_INT_MASK_CLOCK should be removed in someway
     93 
     94     o	fix kernel start address
     95 
     96     o	merge new wscons
     97 
     98     o	parse ARC BIOS configuration information and use it
     99 
    100     o	omit __BROKEN_CONFIG_UNIT_USAGE
    101 
    102     o	omit __SWAP_BROKEN in <mips/types.h>
    103 
    104     o	fix implementation of DELAY(), clean up clock implementation
    105 
    106     o	asc.c scsi clock/NCR53CF94 handling clean up
    107 
    108     o	if_sn.c ether address handling clean up
    109 
    110     o	com_lbus.c clock handling clean up
    111 
    112     o	implement NCR 53c700(?) driver for NEC RISCserver 2200
    113 	based on amiga siop driver?
    114 
    115     o	audio driver
    116 
    117     o	use MI driver
    118 
    119 	-   use MI ncr53c9x driver instead of home grown asc
    120 
    121 	-   use MI bha driver instead of home grown btl
    122 
    123 	-   make sonic driver MI, and share it with mac68k, newsmips/apbus
    124 
    125 	-   make fd driver MI, and share it with i386
    126 	    (contact christos about MI fd driver)
    127 
    128 	-   make pccons MI, and share it with i386,
    129 	    or simply eliminate pccons
    130 
    131 	-   LKM
    132 
    133     o	ARC boot device name -> NetBSD root device conversion
    134 
    135     o	and missing MI devices
    136 		ses?, lkm, vcoda, ...
    137 
    138     o	way to specify serial console
    139 
    140     o resolve "XXX"
    141 
    142 Lots of other things.....
    143