TODO revision 1.36
1 /* $NetBSD: TODO,v 1.36 2021/02/27 18:29:38 palle Exp $ */ 2 3 Things to be done: 4 5 common: 6 - make %g6 point to curcpu 7 - make %g7 point to curlwp 8 - change run-time checks for cpu type to function pointers 9 10 sun4u: 11 - GENERIC.UP kernel hangs on v445 (missing interrupt?) 12 13 sun4v: 14 - current status 15 T5 ldom with 2 VCPU and 4GB: 16 The kernel boots and starts userland when booting miniroot.fs. 17 The sysinst tool starts properly and is functional. 18 Installation is not possible, since no driver currently exists 19 for the ldom network interfaces, but openbsd vnet and vsw drivers are 20 possible to integrate and this work is progressing. 21 OpenBoot (properly trashed registers). 22 T2000 ldom with 8 VCPU and 4GB: 23 On this platform it crashes in /sbin/init doing an access() call where %o0 is corrupted (zero) 24 - 64-bit kernel support 25 - 32-bit kernel support 26 - libkvm 27 - ofwboot: tlb_init_sun4v() hardcodes number of slots to 64 28 - locore.s: sun4v_datatrap missing implementation for trap level 1 29 - check build without SUN4V defined 30 - replace relevant references to %ver with GET_MAXCWP 31 - pmap_mp_init(): sun4v missing handling 32 - replace constructs like "wrpr %g0, PSTATE_KERN, %pstate" with NORMAL_GLOBALS 33 - replace constructs line "wrpr %g0, PSTATE_INTR, %pstate" with ALTERNATE_GOBALS 34 - sun4v tsb no need to lock... per cpu... anyway... 35 - ci_tsb_desc->td_ctxidx: -1 or 1? 36 - MP support - currently bypassed in pmap_bootstrap() for sun4v 37 - vpci.c/vpcivar.h: cleanup FIXMEs 38 - interrupts not handled properly (com at ebus only...) 39 - mpt(4) complains: mpt0: Phy 0: Link Status Unknown 40 - man pages for drivers imported from OpenBSD lke vpci, vbus, cbus, vdsk, ldc etc. 41 - vdsk and ldc drivers: code maked with OPENBSD_BUSDMA - make the bus_dma stuff work properly 42 - vbus.c: handle prom_getprop() memory leaks 43 - locore.s: rft_user (sun4v specific manaul fill) - seems to work, but is it good enough (compared to openbsds rft_user? 44 - openfirmware interface 45 bounce handlng could need rework 46 - only #if __arch64__ independent for SUN4V, as OF for all ultrasparcs is 47 plain 32bit code too (and this #if only because the point below becomes 48 always false otherwise) 49 - only bounce if the address is >= 1UL<<32 50 - perhaps apply the lock to OF calls in general, not only the bounce'd ones 51 (in other words: let's protect the OF context with the lock, not 52 only the bounce buffer contents). 53 - Are there any other bounce-worthy OF calls? If so, apply handling 54 to all of them now 55 - consider method used by Solaris/OpenSolaris/Illumos (look for PROM_32BIT_ADDRS in 56 the Illumos codebase) 57 - platforms tested so far: 58 -- qemu sun4v/niagara - reaches the point where the kernel needs to mount a root fs 59 -- T2000 - mmu fault somewhere 60 -- T5 - stickcmpr is not properly initialized ??? 61
Indexes created Tue Sep 30 17:09:57 GMT 2025