TODO revision 1.20 
1$NetBSD: TODO,v 1.20 2005/04/19 12:12:28 tsutsui Exp $
2
3To do list (in some particular order)
4XXX some entries might be obsolete.
5
6    o	sysinst
7	maybe MD fdisk partitioning support is required to load files
8	from FAT partition on ARC BIOS prompt.
9
10    o	install notes
11
12
13    o	use MI driver
14
15	-   make fd driver MI, and share it with other ports
16	    (contact christos about MI fd driver)
17
18	-   use MI bha driver instead of home grown btl
19	    XXX needs fixes of DESKstation support
20
21    o	VXL framebuffer support (Magnum, RISCstation 2200)
22
23    o	com_jazzio.c
24	-   clock handling clean up (obtain from ARC BIOS)
25	-   fifo disabling may be only needed on some Magnum?
26
27    o	remove pccons and switch to wscons completely
28	(XXX what's the problem to remove pccons?)
29
30    o	AD1848 audio support
31
32    o	missing MI devices
33		ses?, vcoda, ...
34
35
36    o	Xserver
37
38	-   VXL		Magnum, RISCstation 2200
39	-   vga/S3	PICA, Image RISCstation - OpenBSD's?
40	-   vga/cirrus	RISCserver 2200, Express5800/240 R4400 EISA
41	-   vga/???	DESKstation Tyne, rPC44
42	-   TGA		RISCstation 2250, Express5800/230 R4400 PCI
43
44    o   Find out why bitmap load to S3-928 flashes screen. (X server)
45	Know why (enable linear mode). Need S3 info.
46
47
48    o	repair DESKstation support
49	-   requires bounce buffer bus_dma for Tyne
50		XXX - too small bounce buffer size (128KB)
51
52    o	Olivetti M700 support
53
54    o	NEC Express5800/230 R10000 PCI (NEC-J95) support
55	(needs MI R10000 support)
56
57    o	SNI RM200PCI/RM300/RM400/RM600 support
58
59
60    o	parse ARC BIOS configuration information and use it
61
62    o	increase MAXPHYS to 64KB
63	(XXX why is it limited to 32KB?)
64
65    o	fix kernel start address
66	(maybe requires bootloader support)
67
68    o	allocate PICA_TL_BASE dynamically
69
70    o	remove inb/outb
71
72    o	remove UADDR
73
74    o	fix mem_clusters[] usage.
75
76    o	intrcnt[] name cleanup, use MI evcnt(9)
77
78    o	test and merge soren's clean up about proc0.p_addr.
79
80    o	redesign interrupt handler framework.
81	i/o bus devices should have sane IPL, but currently doesn't.
82
83	also, current MIPS interrupt handler has overblocking and
84	other problems as follows:
85
86	-   SR_INT_IE should be enabled before calling hardclock().
87	    Since this is not done currently, spllowersoftclock()
88	    on hardclock() doesn't have effect, and softclock() is
89	    handled with all interrupt disabled in this case.
90		-> overblocking, possibly causes missing hardclock()
91
92	-   MIPS3_CLKF_BASEPRI() doesn't work correctly,
93	    when MIPS_INT_MASK_5 (== MIPS_INT_MASK_CLOCK) is disabled.
94		-> micro optimization on hardclock() doesn't work.
95		   but currently this may make hardclock() latency better
96		   due to above SR_INT_IE problem.
97	    s/MIPS_INT_MASK/MIPS3_INT_MASK/ makes this work, although tricky.
98
99	-   if (ipending & INT_MASK_REAL_DEV) == 0,
100	    softnet() and softclock() are handled with all interrupt disabled.
101		-> overblocking, possibly causes missing hardclock()
102
103	-   `netisr' handling in netintr() implies potential race condition.
104	    The access to `netisr' should be protected by splnet().
105	    Currently this is not real problem due to above overblocking.
106
107	-   INT_MASK_REAL_DEV should be removed
108
109	-   make CLKF_INTR() work.
110
111    o	it is better to always disable MIPS_INT_MASK_CLOCK.
112	those are the points which should be fixed:
113	mips_idle: li t0, (MIPS_INT_MASK | MIPS_SR_INT_IE)
114	machdep.c: curpcb->pcb_context[11] = MIPS_INT_MASK | MIPS_SR_INT_IE;
115	spl0()
116	splnone()
117
118	-   MIPS_INT_MASK_CLOCK should be removed in someway
119
120    o	XXX at least 2000/06/07 version is already quite unstable
121	on PICA and NEC Image RISCstation. (but almost OK on Magnum)
122	Userland commands dumps	core randomly.
123	This version is before _MIPS_PADDR_T_64BIT changes
124	and MIPS3_TLB_WIRED_UPAGES changes.	
125
126	"vm_page_zero_enable = FALSE" makes this problem disappeared.
127	(vm_page_zero_enable = FALSE by default on all archs w/ UBC, now)
128
129	currently, page zero in the idle loop is also disabled on
130	untested platforms like DESKstation rPC44/Tyne and SNI for safety.
131
132	    XXX what's the current status of uvm_pageidlezero()?
133
134
135    o	resolve "XXX"
136
137
138(following entries might be MI MIPS items)
139
140    o	Move the RO and WIRED attribute from the pte to the pv table.
141	This saves four instructions in the tlb miss handler.
142
143    o   Can we have 32 double registers?
144
145    o	64bit kernel/userland
146
147    o	fix implementation of DELAY(), clean up clock implementation
148
149    o	omit __SWAP_BROKEN in <mips/types.h>
150
151    o	clean up ALEAF/NLEAF/NON_LEAF/NNON_LEAF in userland.
152
153Lots of other things.....
154