area. These functions are designed to improve performance of large
copyin/copyout operations by mapping the user page in to the kernel
address space and using bcopy(), rather then copying across protection
boundaries.
XXX This doesn't work yet -- the way it's called doesn't obey C calling
XXX conventions. That will be fixed soon.
- copypage() -- a single page-aligned NBPG-byte copy.
- zeropage() -- a single page-aligned NBPG-byte zero.
These functions don't play around with alignment, etc. Their use
causes a measureable performance improvement in pmap_copy_page()
and pmap_copy_page().
A few m68k ports already had copypage() in their locore.s. It has
been moved here so it can be shared.
- Synch the "clean" rule with the i386 port's.
- Add a commented-out rule for generating assym.h w/ the new genassym.sh.
Some slight changes need to be made to genassym.h to make it work with
m68k, so we can't use it just yet.
barring any more little things people want added ...]
New features:
* progressmeter is now asynchronous, so "stalled" transfers can be
detected. "- stalled -" is displayed instead of the ETA in this case.
When the xfer resumes, the time that the xfer was stalled for is
factored out of the ETA. It is debatable whether this is better than
not factoring it out, but I like it this way (I.e, if it stalls for 8
seconds and the ETA was 30 seconds, when it resumes the ETA will still
be 30 seconds).
* verbosity can be disabled on the command line (-V), so that in auto-fetch
mode the only lines displayed will be a description of the file, and
the progress bar (if possible)
* if the screen is resized (and detected via the SIGWINCH signal), the
progress bar will rescale automatically.
Bugs fixed:
* progress bar will not use the last character on the line, as this can
cause problems on some terminals
* screen dimensions (via ioctl(TIOCWINSZ)) should use stdout not stdin
* progressmeter() used some vars before initialising them
* ^D will quit now. [fixes bin/3162]
* use hstrerror() to generate error message for host name lookup failure.
* use getcwd instead of getwd (it should have been OK, but why tempt fate?)
* auto-fetch transfers will always return a positive exit value upon failure
or interruption, relative to the file's position in argv[].
* remote completion of / will work, without putting a leading "///".
This is actually a bug in ftpd(1), where "NLST /" prefixes all names
with "//", but fixing every ftpd(1) is not an option...
#defining MALLOC/FREE to use malloc()/free() if DIAGNOSTIC (not DEBUG)
since DIAGNOSTIC is what enables freelist consistency checking, and
don't force the definition of KMEMSTATS if DEBUG is defined (there's no
reason to, since users can do it themselves if they want it).
pass the block number of the boot partition to the kernel. The kernel
will use the block number to determine which disk device the kernel was
booted from and set the boot device based on that instead of the old
"generic" method.
thorpej