touched any user-space address recently. This is efficient
for things that stay in the kernel for a while, waking up
to handle some I/O then going back to sleep (i.e. nfsd).
If and when such a process returns to user-mode, it will
fault and be given a real context at that time.
This also makes context switch faster, because all we need
to do there for the MMU is slam the context register.
since today, they will have the same size as the on-the-wire-packet on each
architecture.
Problem was reported by George Harvey for the m68k architecture.
+ * stor-layout.c (layout_record): Ignore STRUCTURE_SIZE_BOUNDARY if
+ we are packing a structure. This allows a structure with only
+ bytes to be aligned on a byte boundary and have no padding on a
+ m68k.
Add more #ifdef DEBUG to target-prefixing.
* Fix niggle in creating a new resolv.conf: ctime()'s result ends in '\n',
so don't include one after the %s for the timestamp.
* Change upgrade: add the same ``disk stuff done'' message used in install,
so the user knows what's going to happen.
(is a more tailored message useful?)
* Make the ring buffer size and water marks patchable, and allocate the buffer
separately.
* Do the ttymalloc() at attach time.
* Reorganize the receive buffer so the status and data pair are next to each
other. This is slightly faster.
* Make sure we actually do turn off interrupts in comclose() if we have DDB
configured and it's not the console. (D'oh!!!!)
* When we exhaust the current transmit run, turn off transmit interrupts in
comintr(), so we're fairly sure we don't get another one.
* Nuke the silly lsrmap[] idea; it's slower in the normal case.
* Cache the l_rint pointer in the soft interrupt routine.
* Carrier detect (TS_CARR_ON) is based on the actual DCD bit, even if it's
being ignored.
* Set TS_WOPEN early on in zsopen().
* Don't disable interrupts on the console during close if we have DDB.
Inert changes:
* Don't handle ZS_HWFLAG_NO_DCD here; the frontend does it.
* Deprecate `register'.
* Use SET(), CLR(), and ISSET().
More performance changes:
* Rototill receive handling; use a backpressure mechanism to prevent livelock.
* Output silo/ibuf overflow warnings at most once per minute, from a callout.
* When we exhaust the current transmit run, turn off transmit interrupts in
zstty_txint(), so we're fairly sure we don't get another one.
* Make the ring buffer size and water marks patchable, and allocate the buffer
separately.
* Do the ttymalloc() at attach time.
* Reorganize the receive buffer so the status and data pair are next to each
other. This is slightly faster.
* Make sure we actually do turn off interrupts in comclose() if we have DDB
configured and it's not the console. (D'oh!!!!)
* When we exhaust the current transmit run, turn off transmit interrupts in
comintr(), so we're fairly sure we don't get another one.
* Nuke the silly lsrmap[] idea; it's slower in the normal case.
* Cache the l_rint pointer in the soft interrupt routine.
--entries / remove kernel entry from locore / reorganize vector entry
--code. Enables access to stack frames transitively w.r.t. palcode
--vectors, e.g. upward traceback works, inverting (possibly several)
--kernel vectors. Until now, all trackbacks ended at the first-reached
--instance of trap(), which was totally useless as there is no mystery
--to trap->panic->cpu_reboot
* Correct arg typo in mv_within_target_or_die().
* Add new path-prefixing entrypoints to fopen(), collect(), and do mounts.
* Use the above. MI code is now clean of explicit references to /mnt.
* Lint: add `const' to collect()'s pathname arg.
include <stdio.h> in factor.c, now that defs.h uses FILE*.
* Put newfs-and-mount code into a separate function, in case
we want to support non-ffs filesystems.
* Check for the currently-mounted root, and don't newfs or mount it.
Assumes md code will also avoid changing that partition.
* renames of files from one pathname within the target to another
(e.g., mv_within_target_or_die("/etc", "/etc.old");
* check to see if a partition name like "sd0a" is the current root
* Duplicate a file from the current root into the target root
(for copying /netbsd from RAMdisk into the target).
A no-op if the root is the install target).
Start using these where appropriate.
Change net.c to avoid losing any information when updating network
config files: where possible, do appends to files that might have more
info than we got from the user (e.g., /etc/hosts.)
Where possible, add comment saying file was created/modified by sysinst.
destination register bit pattern with 1.0), which automatically provides
corner case handling.
Missing ftwotox emulation originally reported by Norman Mackenzie in PR 4237,
but he proposed a different implementation.
- Map in all physical memory first in system space. This reduces
pmap_copy_page() and pmap_zero_page to 3 resp. 2 instructions.
- Have fized position user page tables. Makes the pv_table smaller
and all reverse references faster (and simpler).
- Remove the wiring code. Nobody doesn't even know what a DR32 is anymore.
- Simulate page reference bit by setting page invalid, as suggested by
Rich Draves in a paper for 1991 Mach Usenix Symposium.
This reduced the time spent in the pmap module to between 70-75% of
the previous; and made process startup _much_ faster.
- Map in all physical memory first in system space. This reduces
pmap_copy_page() and pmap_zero_page to 3 resp. 2 instructions.
- Have fized position user page tables. Makes the pv_table smaller
and all reverse references faster (and simpler).
- Remove the wiring code. Nobody doesn't even know what a DR32 is anymore.
- Simulate page reference bit by setting page invalid, as suggested by
Rich Draves in a paper for 1991 Mach Usenix Symposium.
This reduced the time spent in the pmap module to between 70-75% of
the previous; and made process startup _much_ faster.
mycroft