Home | History | Annotate | Line # | Download | only in arc
TODO revision 1.19 
      1 $NetBSD: TODO,v 1.19 2003/05/25 15:46:15 tsutsui Exp $
      2 
      3 To do list (in some particular order)
      4 XXX some entries might be obsolete.
      5 
      6     o   Boot. Standalone boot program to load ELF kernels instead of
      7 	booting ECOFF kernels directly.
      8 	(maybe sgimips/stand would help, but annoying buggy ARC BIOS)
      9 
     10     o	sysinst
     11 	maybe MD fdisk partitioning support is required to load files
     12 	from FAT partition on ARC BIOS prompt.
     13 
     14     o	install notes
     15 
     16 
     17     o	use MI driver
     18 
     19 	-   make fd driver MI, and share it with other ports
     20 	    (contact christos about MI fd driver)
     21 
     22 	-   use MI bha driver instead of home grown btl
     23 	    XXX needs fixes of DESKstation support
     24 
     25     o	VXL framebuffer support (Magnum, RISCstation 2200)
     26 
     27     o	com_jazzio.c
     28 	-   clock handling clean up (obtain from ARC BIOS)
     29 	-   fifo disabling may be only needed on some Magnum?
     30 
     31     o	remove pccons and switch to wscons completely
     32 	(XXX what's the problem to remove pccons?)
     33 
     34     o	AD1848 audio support
     35 
     36     o	missing MI devices
     37 		ses?, vcoda, ...
     38 
     39 
     40     o	Xserver
     41 
     42 	-   VXL		Magnum, RISCstation 2200
     43 	-   vga/S3	PICA, Image RISCstation - OpenBSD's?
     44 	-   vga/cirrus	RISCserver 2200, Express5800/240 R4400 EISA
     45 	-   vga/???	DESKstation Tyne, rPC44
     46 	-   TGA		RISCstation 2250, Express5800/230 R4400 PCI
     47 
     48     o   Find out why bitmap load to S3-928 flashes screen. (X server)
     49 	Know why (enable linear mode). Need S3 info.
     50 
     51 
     52     o	repair DESKstation support
     53 	-   requires bounce buffer bus_dma for Tyne
     54 		XXX - too small bounce buffer size (128KB)
     55 
     56     o	Olivetti M700 support
     57 
     58     o	NEC Express5800/230 R10000 PCI (NEC-J95) support
     59 	(needs MI R10000 support)
     60 
     61     o	SNI RM200PCI/RM300/RM400/RM600 support
     62 
     63 
     64     o	parse ARC BIOS configuration information and use it
     65 
     66     o	increase MAXPHYS to 64KB
     67 	(XXX why is it limited to 32KB?)
     68 
     69     o	fix kernel start address
     70 	(maybe requires bootloader support)
     71 
     72     o	allocate PICA_TL_BASE dynamically
     73 
     74     o	remove inb/outb
     75 
     76     o	remove UADDR
     77 
     78     o	fix mem_clusters[] usage.
     79 
     80     o	intrcnt[] name cleanup, use MI evcnt(9)
     81 
     82     o	test and merge soren's clean up about proc0.p_addr.
     83 
     84     o	redesign interrupt handler framework.
     85 	i/o bus devices should have sane IPL, but currently doesn't.
     86 
     87 	also, current MIPS interrupt handler has overblocking and
     88 	other problems as follows:
     89 
     90 	-   SR_INT_IE should be enabled before calling hardclock().
     91 	    Since this is not done currently, spllowersoftclock()
     92 	    on hardclock() doesn't have effect, and softclock() is
     93 	    handled with all interrupt disabled in this case.
     94 		-> overblocking, possibly causes missing hardclock()
     95 
     96 	-   MIPS3_CLKF_BASEPRI() doesn't work correctly,
     97 	    when MIPS_INT_MASK_5 (== MIPS_INT_MASK_CLOCK) is disabled.
     98 		-> micro optimization on hardclock() doesn't work.
     99 		   but currently this may make hardclock() latency better
    100 		   due to above SR_INT_IE problem.
    101 	    s/MIPS_INT_MASK/MIPS3_INT_MASK/ makes this work, although tricky.
    102 
    103 	-   if (ipending & INT_MASK_REAL_DEV) == 0,
    104 	    softnet() and softclock() are handled with all interrupt disabled.
    105 		-> overblocking, possibly causes missing hardclock()
    106 
    107 	-   `netisr' handling in netintr() implies potential race condition.
    108 	    The access to `netisr' should be protected by splnet().
    109 	    Currently this is not real problem due to above overblocking.
    110 
    111 	-   INT_MASK_REAL_DEV should be removed
    112 
    113 	-   make CLKF_INTR() work.
    114 
    115     o	it is better to always disable MIPS_INT_MASK_CLOCK.
    116 	those are the points which should be fixed:
    117 	mips_idle: li t0, (MIPS_INT_MASK | MIPS_SR_INT_IE)
    118 	machdep.c: curpcb->pcb_context[11] = MIPS_INT_MASK | MIPS_SR_INT_IE;
    119 	spl0()
    120 	splnone()
    121 
    122 	-   MIPS_INT_MASK_CLOCK should be removed in someway
    123 
    124     o	XXX at least 2000/06/07 version is already quite unstable
    125 	on PICA and NEC Image RISCstation. (but almost OK on Magnum)
    126 	Userland commands dumps	core randomly.
    127 	This version is before _MIPS_PADDR_T_64BIT changes
    128 	and MIPS3_TLB_WIRED_UPAGES changes.	
    129 
    130 	"vm_page_zero_enable = FALSE" makes this problem disappeared.
    131 	(vm_page_zero_enable = FALSE by default on all archs w/ UBC, now)
    132 
    133 	currently, page zero in the idle loop is also disabled on
    134 	untested platforms like DESKstation rPC44/Tyne and SNI for safety.
    135 
    136 	    XXX what's the current status of uvm_pageidlezero()?
    137 
    138 
    139     o	resolve "XXX"
    140 
    141 
    142 (following entries might be MI MIPS items)
    143 
    144     o	Move the RO and WIRED attribute from the pte to the pv table.
    145 	This saves four instructions in the tlb miss handler.
    146 
    147     o   Can we have 32 double registers?
    148 
    149     o	64bit kernel/userland
    150 
    151     o	fix implementation of DELAY(), clean up clock implementation
    152 
    153     o	omit __SWAP_BROKEN in <mips/types.h>
    154 
    155     o	clean up ALEAF/NLEAF/NON_LEAF/NNON_LEAF in userland.
    156 
    157 Lots of other things.....
    158