to schedule clock interrupts at a fixed interval, rather scheduling
the next one based on the time of the arrival/servicing of the previous
clock interrupt. Also, pick up a trick from the sbmips port to convert
a division in ip22_clkread to a multiplication, since those are much
cheaper -- the details of that are described in Simon's commit (see
Message-Id: <20020306073437.1D2A8B004@cvs.netbsd.org>). Thanks to
Jason Thorpe and Dominic Sweetman's "See MIPS Run" (where I found
mention of this very subject while looking for something totally un-
related! 8-) for the clue about the source of the timekeeping problems.
For the IP32, where we have no clock-calibration code yet, use the CPU
frequency provided by ARCS instead; it beats a hard-coded value!
As an added bonus, most of the CPU-clock related stuff is now collected
together in cpu_info_store, rather than as a collection of unorganized
global variables.
Obsolete NBUILDJOBS; build.sh just passes -jN through to make(1),
which inherits it cooperatively through the build tree. Fix
documentation so that it's shown to be deprecated.
If you use build -jN, please save full build logs so that errors due
to missing dependancies can be analyzed and corrected.
- move guts of distrib/Makefile.inc to distrib/common/Makefile.distrib
(fixes problem caused by implicit include of ../Makefile.inc in certain
submake conditions triggered by makefiles not yet in tree)
- removed mkdir of ${RELEASEDIR}/*; rely upon "snap_pre" target of
etc/Makefile to create all the release directories
- renamed RELINSTALL to RELEASE_INSTALL
- renamed FLOPPYINSTDIR to FLOPPY_RELEASEDIR
- renamed MDSETDIR to MDSET_RELEASEDIR
- removed ITARGET
- move release target from top level to appropriate subdirectory
- ensure release target has correct depends
- replace miniroot's IMAGE_MD_POST with common/Makefile.image IMAGEPOSTBUILD
- Makefile.image: add realall: ${IMAGE}
by default, and can be enabled by adding the SOSEND_LOAN option to your
kernel config. The SOSEND_COUNTERS option can be used to provide some
instrumentation.
Use of this option, combined with an application that does large enough
writes, gets us zero-copy on the TCP and UDP transmit path.
userlevel; this is necessary due to the 601, unlike other 6xx, having
no concept of separated Valid_user vs. Valid_supervisor for BATs.
* When crossing the kernel/userlevel boundary, have platform-provided
hooks set up the two fixed BAT entries, and possibly additional
segment registers to redeem the 601's BAT limitations.
Both of the above are only built if the $MACHINE provides these hooks,
sparing others the pain.
transmit and receive descriptor rings is limited to 256 descriptors.
So, set the if_snd queue length to 256 to let the upper layers queue
lots of packets, and let the driver handle up to 32 of them at a time.
(We should probably make this change to most Ethernet drivers, since
it actually saves some resources.)
* Increase the number of Tx DMA segments from 8 to 16.
* Clean up the way we count "how many times did I get a packet with N
DMA segments".
* Add a missing htole32() in wm_tx_cksum().
* Don't set both RS and RPS in the last Tx descriptor of a packet; just
use RS.
* Add some more information to the watchdog message.
