Recent additions to NetBSD/sgimips

Rafal K. Boni has added a lot of new support to NetBSD/sgimips that has been committed to NetBSD-current. Most notably is new support for the IP22 (Indigo2, Indy, Challenge S). A list of all the major new additions is as follows:

• Support for the IP22 (Indigo2, Indy, Challenge S)
• Support for the HPC I/O peripheral ASIC. There is where SCSI, Ethernet, etc. live on the Indigo2 and Indy
• Support for the Seeq 8003 Ethernet attached to the HPC ASIC, found on the Indigo2 and Indy
• Support for the Dallas 1286 clock found on the Indigo2 and Indy
• IOC1/2 memory map
• Support for the Indigo2/Indy IMC (this is that to which all other devices are attached)
• Support for the SCC serial ports on Indigo2 and Indy
• Support for configuring an SCC console
• Generic software interrupt implementation, derived from NetBSD/mipsco
• Basic EISA support for the Indigo2
• Dynamic system configuration support, derived from NetBSD/pmax

An example kernel configuration file for Rafal's Indigo2 can be found in sys/arch/sgimips/conf/TEAL. [0 comments]

 

Port page created (sun2)

NetBSD/sun2 now has a port page on www.NetBSD.org. [0 comments]

 

(vax) Two new drivers committed

Anders Magnusson has committed a driver for the Interlan NI1010 Ethernet Controller (il) and a driver for the DEC DDCMP DMC-11/DMR-11 interface (dmc) to NetBSD/vax -current. Both drivers are from 4.4BSD. [0 comments]

 

Support for VAX 4000/705A

Michael Kukat has written support for the VAX 4000/705A. [0 comments]

 

NetBSD/mvme68k now using MI 53c710 driver

NetBSD/mvme68k is now using the MI driver for 53c710 SCSI controller recently committed by Izumi Tsutsui. If you're tracking -current on NetBSD/mvme68k and have you own custom kernel config files for mvme1[67][27] (i.e. all models except mvme147), you'll need to replace any reference to `ncrsc' (the old driver) with `osiop' (the new MI driver). [0 comments]

 

VAX-specific in4_cksum() function repaired

Anders Magnusson has repaired the VAX-specific in4_cksum() function, which, in turn, will improve performance. For example, a VAX3900 with DEQNA (the bad one) peeks just above 500K/s now, around 50% of the cable capacity, which is actually very good. [0 comments]

 

Hardware scrolling for RiscPC/A7000/NC wscons

Reinoud Zandijk has committed code for hardware scrolling for the RiscPC and compatibles. This new code has been tested in different screen resolutions and depths, and its all worked ok so far. It even appears to run faster than the original vidcconsole code. More details are available in Reinoud's announcement in the port-arm32 mail archive. [0 comments]

 

NetBSD/sparc using a pool cache for pmap allocation

Paul Kranenburg has updated NetBSD/sparc to use a pool cache for pmap allocation. The cached pmap structures include a ready-to-use level 1 page table (on sun4m) and MMU context. A simple `fork()' test shows an improvement in spent system time of around 4 percent. [0 comments]

 

NetBSD/news68k switches to ELF

NetBSD/news68k is now using ELF as its native object file format in -current. An ELF snapshot is available in the NetBSD/news68k snapshot area of the ftp server. Also available are instructions for upgrading from this snapshot or from source. [0 comments]

 

Linux compatibility enabled for PowerPC ports

Emmanuel Dreyfus has integrated Linux binary compatibility for PowerPC ports. This means that NetBSD/prep should now be able to run LinuxPPC binaries such as Netscape Communicator. [0 comments]

 

Linux compatibility enabled for PowerPC ports

Emmanuel Dreyfus has integrated Linux binary compatibility for PowerPC ports. This means that NetBSD/bebox should now be able to run LinuxPPC binaries such as Netscape Communicator. [0 comments]

 

Physical memory handling changes

Paul Kranenburg has updated the code that manages physical memory in NetBSD/sparc (see the NetBSD-current versions of pmap.c, machdep.c, and autoconf.c), resulting in these "user visible" changes:

The kernels message buffer size is now 8192 (was 4096) on all platforms under NetBSD/sparc. sun4 users: The full first page is now used, since there should be sufficient sanity checking in the message buffer initialization code to detect possible corruption by boot programs. Testing on a sun 4/490 showed no problems.

sun4m PROMs may allow you to not have a memory bank starting at physical address 0. The NetBSD bootstrap code now also deals with this, so you can plug your memory modules into any slot supported by the machine's PROM.

If you encounter any problems, please drop Paul a note.

[0 comments]