Added hook for pmap_prefault() and use symbolic constant for new third
argument to vm_page_alloc() (vm_fault.c, various)
Changed the way that upages and page tables are held. (vm_glue.c)
Fixed architectural flaw in allocating pages at interrupt time that was
introduced with the merged cache changes. (vm_page.c, various)
Adjusted some algorithms to acheive better paging performance and to
accomodate the fix for the architectural flaw mentioned above. (vm_pageout.c)
Fixed pbuf handling problem, changed policy on handling read-behind page.
(vnode_pager.c)
Submitted by: John Dyson
implemented the ability to limit bufferspace by memory consumed. (vfs_bio.c)
Fixed recently introduced bugs that caused extra I/O to happen in some
cases. (vfs_cluster.c)
Submitted by: John Dyson
Moved various pmap 'bit' test/set functions back into real functions; gcc
generates better code at the expense of more of it. (pmap.c)
Fixed a deadlock problem with pv entry allocations (pmap.c)
Added a new, optional function 'pmap_prefault' that does clustered page
table preloading (pmap.c)
Changed the way that page tables are held onto (trap.c).
Submitted by: John Dyson
page flipping anytime soon and 4k just wastes memory. Performance via the
looback interface will decrease slightly by this change. Idealy, we need
buffers that can have variable sizes, but this requires a rewrite of far
too much code at the moment.
1) Don't clone routes to multicast destinations; there is nothing useful
to be gained in this case.
2) Reduce default expiration timer to one hour. Busy sites will still
likely want to reduce this, but for ordinary users this is a reasonable
value to use.
delete them when the ``parent'' goes away
route.h: add glue to track this to rtentry structure. WARNING WILL ROBINSON!
This will be yet another incompatible change in your route-using binaries.
I apologize, but this was the only way to do it. I took this opportunity
to increase the size of the metrics to what I believe will be the final
length for 2.1, so that when the T/TCP stuff is done, this won't happen
again.
Print out the length of the compiled sequencer program.
aic7xxx.seq:
More optimizations. Replace generic bcopy routine with bcopy_3
and bcopy_4 (ie unroll the loops) since these are the only two cases used.
Initialize SIMODE1 and SXFRCTL1 from the kernel in ahc_init instead of
at each selection/reselection since this is expensive and only needs to
be done once. Condense function returns into previous instruction if possible.
Reorder some sections to kill superflous jumps. These optimizations kill
the ~150k/s penalty adding support for Twin/Wide cards was costing since
the last place in the commaon path of execution where we had to do ugly,
convoluted testing for the type of card in the sequencer has gone away.
Next stop tagged queuing and target mode.
work (mi_switch() counted the last timeslice again but this didn't affect
the exiting process' rusage because the rusage has already been finalized).
Remove stale comment.
the physical device is closed. Previously only the reverse case was handled.
Abuse the cdevsw interface instead of the vfs interface to do this.
Remove unnecessary #includes.
When using cp to copy a file under the following circumstanes:
- original file in on an NFS filesystem
- destination file is on the same NFS filesystem
- the file is less than 8Mbytes in size
- the file is larger than 65536 bytes in size
the cp process can get frozen in device-wait and never wake up (cp uses
mmap() in this case).
A small change to allocbuf() fixes this.
Obtained from:
bios boot block changed to allow booting from both the attached graphics
display and from a serial port. (A specially compiled serial boot block
is no longer necessary.) The boot block should detect the presence or
absence of a keyboard: if there is no keyboard, COM1 is turned into the
console. This simulates the behavior of the Sun boot PROMs. Unplug your
keyboard, attach a terminal to COM1 and you should be ready to go. :)
Obtained from:
sio.c and sioreg.c changed to allow autodetecting the RB_SERIAL flag
passed by the boot blocks so that the kernel can switch to 'serial
console' mode automagically. 'options COMCONSOLE' can still be specified
to force the kernel to always use the serial port as a console.
CONUNIT and CONADDR can also be specified in the kernel config file
if the user wants to shift the console to a different port.
Obtained from:
Added a new flag (RB_SERIAL) which is passed by the boot block to tell
the kernel that it should use a serial port as a console. (The kernel
can still be forced to boot in serial console mode if 'options COMCONSOLE'
is used.)
Put in the much shorter and cleaner version for the calibrate_cycle_counter
for the Pentium that Bruce suggested. Tested here on my Pentium and
it works okay.
sigreturn() sometimes failed for ordinary returns from signal handlers.
Failures of ordinary returns "can't happen" and are badly handled.
"Temporary" fix: allow users to corrupt PSL_RF. This is fairly
harmless. A correct fix would involve saving the old %eflags (and
perhaps the old segment registers) where the user can't get at them.
(original "High Sierra") CD format. I've already implemented this for
1.1.5.1 (and posted to -hackers), but didn't get any response to it.
Perhaps i'm the only one who has such an old CD lying around...
Everything is done empirically, but i had three of them around (from
different vendors), so there's a high probability that i've got it
right. :)
WDTR, and message reject handlers so they don't need to exist in the
sequencer. All three of these cases are not on the critical path, so it
makes little sense to use up precious sequencer ram for them.
(SCSI control block) instead of having the host PIO it down. Also
reimplement WDTR and SDTR optimization to remove code in the sequencer
and place the responsibility of knowing when to initiate SDTR or WDTR
on the kernel driver. This vastly shortens the sequencer program yet
yeilds the same performance.
properly from the beginning:
1) The `kern_devconf' struct should be a part of the driver's
`softc' structure (now it is).
2) The `description' should say what the device actually is,
rather than just giving a model number (now it does).
3) The device should be registered even if the probe fails, so
that it can be reconfigured later.
4) For netifs, the device state should follow the IFF_UP flag.
Other network interfaces should follow this example. (Please?) Eventually
there should be a rundown routine doing the equivalent of setting IFF_UP
off, and perhaps more if warranted.
attempted to check for insecure and fatal eflags and segment
selectors, but missed many cases and got the IOPL check back to
front. The other syscalls didn't check at all.
sys_process.c, machdep.c:
Only allow PT_WRITE_U to write to the registers (ordinary and FP).
psl.h, locore.s, machdep.c:
Eliminate PSL_MBZ, PSL_MBO and PSL_USERCLR. We are not supposed
to assume anything about the reserved bits. Use PSL_USERCHANGE
and PSL_KERNEL instead. Rename PSL_USERSET to PSL_USER.
exception.s:
Define a private label for use by doreti when returning to user
mode fails.
machdep.c:
In syscalls, allow changing only the eflags that can be changed on
486's in user mode (no longer attempt to allow benign IOPL changes;
allow changing the nasty PSL_NT; don't allow changing the i586
bits).
Don't attempt to check all the cases involving invalid selectors
and %eip's. Just check for privilege violations and let the invalid
things cause a trap.
procfs_machdep.c:
Call the ptrace register functions to do all the work for reading
and writing ordinary registers and for single stepping.
trap.c:
Ignore traps caused by PSL_NT being set. Previously, users could
cause a fatal trap in user mode by setting PSL_NT and executing an
iret, and a fatal trap in kernel mode by setting PSL_NT and making
a syscall. PSL_NT was cleared too late and not in enough modes to
fix the problem.
Make all traps in user mode (except T_NMI) nonfatal.
Recover from traps caused by attempting to load invalid user
registers in doreti by restarting the traps so that they appear to
occur in user mode.
---
Fix bogons that I noticed while fixing the above:
psl.h:
Fix some comments.
Uniformize idempotency ifdef.
exception.s, machdep.c:
Remove rsvd[0-14]. rsvd0 hasn't been reserved since the 486 came
out. Replace rsvd0 by `align'. rsvd[0-11] used wrong (magic
non-unique) trap numbers. Replace rsvd[1-14] by rsvd.
locore.s:
Enable alignment check flag on 486's and 586's.
machdep.c:
Use a better type for kstack[].
Use TFREGP() to find the registers.
Reformat ptrace functions from SEF to something closer to KNF.
procfs_machdep.c:
The wrong pointer to the registers got fixed as a side effect.
Implement reading and writing of FP registers.
/proc/*/*regs now work (only) for processes that are in memory.
Clean up comments.
trap.c, trap.h:
Remove unused trap types.
unreachable case label in kdb_trap().
Use the correct case labels in kdb_trap() so that normal ddb entry doesn't
print a message.
Change all printf's to db_printf's. Now you can put a breakpoint at printf,
and ddb entry messages don't spam the syslog output.
Cosmetic:
Use ISPL() instead of magic numbers.
Don't compile the unused function kdb_kbd_trap().
Improve some asms.
Print the arg to Debugger().
need to be moved into the qcollapse and rcollapse routines, but I don't
have time at the moment to make all the required changes...this will do
for now.
sense retrieval code that messed up CDROM devices. This code will also
responde correctly to SDTR and WDTR messages from devices that start a
negotiation sequence.
protected drive at open() time has been *totally bogus*! The guy who
submitted it didn't understand all the implications of calling
set_motor(), and the `who' who included the patch into the tree did it
blindly... Pleeeeze, don't commit code to this driver unless you are
really going to understand what it does! This one caused me to pull
out even more hears, and those who know me do know that i ain't got
too many o'them. :-)
1) Firewall is not subdivided on forwarding / blocking chains
anymore.Actually only one chain left-it was the blocking one.
2) LKM support.ip_fwdef.c is function pointers definition and
goes into kernel along with all INET stuff.
firewall should *NOT* be compiled into kernel.
Then it can be loaded.This is misc module but i'v
got no problemms with it,so shouldn't you i suppose..
BTW this is very stupid to have one module in CVS
for ALL lkm's...
No kernel config options anymore besides keyboard language layout.
Virtual consoles are now dynamically allocated, no NCONS anymore.
Software cursor blinking/nonblinking.
Visual bell for laptops (don't beep at meetings :-).
Cursor/bell default type setable via config "flags" instead of as defines.
Cursor/bell type setable via ioctl's.
New video modes 80x30 80x60 for some laptops, and those with multisync monitors.
Scroll-lock history (length currently fixed at 100 lines).
Lots of cleanups, some only commented out for now (will goaway soon).
Support for new features in vidcontrol/kbdcontrol.
Updated manpages.
being cleared in some cases for vnode backed objects; we now do this in
vnode_pager_alloc proper to guarantee it. Also be more careful in the
rcollapse code about messing with busy/bmapped pages.
now marked busy as long as it's being in non-reset state, and the
drives are busy as long as at least one instance is open.
Also reformat everything to fit into 80 columns again.
Changed my mind wrt. error reporting for a write-protected drive and
an open() with write intent; ENXIO has been too weird, now return EIO.
Some portions of the code need to be rewritten to use tprintf()
instead of simple printf()'s, so the messages will also appear on the
session terminal, however.
Wangtek PC-02 QIC-02 controller with Cipher 150MB tape drive
Any attemp to open /dev/rwt0 hangs.
The following patch to /sys/i386/isa/wt.c is derived from Mikael
Hybsch's wt driver.
Submitted by: Terry Lee <terry@uivlsi.csl.uiuc.edu>
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
Reenabled "SCIOCOMAND" ioctl.
Restructured so low level drivers can easily request retries.
Added preliminary fixed SCSI devices (should be revisited before 2.1)
Added "ssc" device that can have its' (HBA, ID, LUN) set via ioctl.
shifting. Also correct the original code as Garrett noticed it in mail.
Leave the mishandled code in to use it later if future versions of gcc
are correct. The code was part of the calibrate_cyclecounter routine to
get the speed of the pentium chip.
floppy driver (or in the hardware?). It turned out to be caused by
spurious interrupts, right after an FDC reset.
Also major cleanup in the low-level structure, there are now functions
performing error-checks for the FDC I/O.
Submitted by: (mostly) Peter Dufault <dufault@FreeBSD.org>
It used to do this by setting a global <Yuck>. Now we set th VNINACT
flag in the vnode to force a skip of ufs_inactive.
Sorry for missing this file in my last commit folks.
Index: vnode.h
===================================================================
RCS file: /usr/cvs/src/sys/sys/vnode.h,v
retrieving revision 1.14
diff -c -r1.14 vnode.h
*** 1.14 1994/11/14 13:51:53
--- vnode.h 1994/12/03 01:06:27
***************
*** 116,121 ****
--- 116,122 ----
#define VALIASED 0x0800 /* vnode has an alias */
#define VDIROP 0x1000 /* LFS: vnode is involved in a directory op */
#define VVMIO 0x2000 /* VMIO flag */
+ #define VNINACT 0x4000 /* LFS: skip ufs_inactive() in lfs_vunref */
/*
* Vnode attributes. A field value of VNOVAL represents a field whose value
ordering that can prove fatal during large batches of deletes, but this
is much better than it was. I probably won't be putting much more time
into this until Seltzer releases her new version of LFS which has
fragment support. This should be availible just before USENIX.
no more DOS boots to start it up.
Simply did a localized nuke of the OUTB macro in this file. This is
a kludge, since it seems it may actually be necessary in other GUS
files (tbd).
Thanks to: Amancio Hasty & Ken Hornstein
correct console number for the VT_WAITACTIVE ioctl. Invalid console numbers
caused waiting on an invalid pointer.
Use bcopyw() instead of move_up() and move_down(). bcopyw() handles
overlapped copies and should be faster. Actually use bcopy(). bcopy()
is slightly faster if video memory is 16-bit and about twice as fast if
it is 32-bit. bcopy() is said to fail on someGA's, but syscons already
depends on it working for other accesses to video memory.
If nd->swap_nblks is zero in nfs_mountroot(), then the system
comes up without initializing swapdev_vp to an actual vnode pointer.
The swap pager assumes a non-NULL value for swapdev_vp.
The fix is to try initializing local swap if no NFS swap space
is specified.
Remove bogus input operands for fnsave(), fnstcw() and fnstsw().
Change all fwait's to fnop's. This might help avoid hardware bugs.
Wait after fninit with an fnop. This should be safer now.
Fix some spelling and formatting errors.
Use natural sizes for control and status words (u_short, promotes to int).
Don't clobber the SWI_CLOCK_MASK bits in npx0_imask when using IRQ13.
Set the devconf state correctly (always busy, if configured). Improve
code for npx_registerdev() a little (gcc can't keep id->id_unit in a
register for some reason). Don't register a nonexistent npx device.
Print a useful message in npxattach() again (delete references to errors
and not the whole message). Don't print "387 emulator" if there is no
emulator in the kernel.
Use %p for pointers in error messages.
Don't clobber the FPU state when there is an FPU exception. Just clear
the exception flags (after saving the flags as before). This allows
debuggers and SIGFPE handlers to look at the full exception state.
SIGFPE handlers should normally return via longjmp(), which restores a
good FPU state (as before). Returning from a SIGFPE handler may leave
the FPU in the wrong state (as before).
Clear the busy latch _after_ clearing the exception flags so that there
is less chance of getting a bogus h/w interrupt for a control operation.
Clear the saved exception status word when the next FPU instruction is
excuted so that it doesn't stick around until the next exception.
Clear the busy latch after fnsave() in npxsave() in case it was set when
npxsave() was called.
values for syncronous negotiation. The 284x series adaptors can now be
supported without the Bios being enabled. If you disable the Bios on the
274x series adaptors, all configuration parameters revert to the default
since there is no way to retrieve them.
You can now sling 14 devices off of a 274xT. In the process of adding
twin channel support, I removed all evident restrictions on supporting
Wide channeled devices, but I do not have a Wide controller to test them
on.
aic7770_seq.h, the pre-compiled header, is no longer needed since config
handles this dependancy.
- /sys/i386/i386/swapgeneric.c is just plain broke. But fear not, for I
have unbroken it. One thing that swapgeneric.c does is walk through the
list of configured devices searching for a boot device. The only easy
way to accomplish this in 2.0 is to use Garret Wollman's kern_devconf
stuff. *BUT*, the head of the kern_devconf linked list (dc_list) is declared
static in /sys/kern/kern_devconf.c. This means that swapgeneric.c can't
see it at link time. I had to remove the 'static' keyword to get around
this little problem. I hope this doesn't break anything anywhere.
*Furthermore,* there's a small matter of making the call to setconf()
in swapgeneric.c disappear when 'config kernel swap generic' isn't used.
You could change /sbin/config to create a dummy setconf() function in
swapkernel.c, but that seems messy somehow. (It's also someting of an
'it isn't broken, why are you fixing it' situation.) My solution was to
do what the NetBSD people did and put an #ifdef GENERIC around the call
to setconf(). If your kernel is called GENERIC or you define 'options
GENERIC,' then you can use 'config kernel swap generic' and it'll work.
That aside, the upshot is that: a) swapgeneric.c actually works, and
and b) the -a boot flag now works as well. If you boot with -a, as in
"Boot: wd(0,a)/kernel -a" you will be presented with a 'root device?'
prompt after the autoconfig phase, at which point you can specify what
device you want mounted as root. Regrettably, you can't specify an NFS
filesystem. Yet. Three files are affected: /sys/i386/i386/swapgeneric.c,
/sys/i386/i386/autoconf.c and /sys/kern/kern_devconf.c.
Submitted by: wpaul
- /sys/i386/isa/if_ed.c doesn't quite know how to deal with SMC EtherEZ
ethernet cards. The EtherEZ looks just like the Elite Ultra, except it
has only 8K of shared memory. The only way to have it properly detected
is to zero and test a few bytes of memory just about the first 8K region.
If it clears properly, it's an Elite Ultra, otherwise it's an EtherEZ.
I've also got an EtherEZ patch for netboot (Makefile, ether.c and ether.h).
- /sys/i386/isa/syscons.c wraps at the next to the last column rather than
the last column, like it should. You don't really notice this unless you
use certain programs that write all the way out to, say, the 80th column,
like VMSmail. Along with a one-line fix for this are some changes to
implement a non-blinking cursor. Put 'options "NOBLINK_CURSOR"' in your
config file and give it a try. :)
Submitted by: wpaul
Would you please commit this two-line patch to /sys/i386/isa/b004.c
(the Transputer driver) so that it at least compiles under 2.x
Haven't tried if the driver is working properly, but a kernel with
compiled-in driver has been running for two days now with no apparent
problems.
Submitted by: luigi
This is new version of Seagate ST01/02, Future Domain TMC-885, TMC-950
SCSI driver for FreeBSD. I started from the 2.0R version and mostly
rewrote it. New features are:
1) New probe algorithm. Old driver read the BIOS region of the adapter
memory and find the copyright string. The problem was in the BIOS itself:
it conflicted with IDE disks. The solution was to unplug it and
make the probe algorithm to work without it.
2) Proper timeout handling in numerous places where the driver
polls waiting for some event.
3) Assembler flagments added in critical places, mostly for data transfer
to of from the target. It was possible to make it faster,
but at the price of decreasing reliability.
4) Target-dependent delays when waiting for REQ deassert event.
Some devices seem to be slower (CD-ROMS, some tape drives),
and some seem to be too fast (disks). The driver tests the REQ
deassert timeout for each target and then uses it for polling.
5) Device flags added for SCSI parity control and sense request
priority control.
6) Generic cleanup, after which the driver became much more readable
(at least by me:).
7) Target data parity error logging is limited to avoid log file overflow.
8) Manual page added.
Submitted by: serge
Move definition of `stat_imask' to clock.c.
clock.c:
Rename `rtcmask' to `stat_imask' and export it. Rename `clkmask' to
`clk_imask' for consistency.
Only calculate TIMER_DIV(hz) once.
Merge debugging and "garbage" code to produce debugging code and format the
output better.
Make writertc() static inline and use it everywhere. Now all accesses to
the clock registers go through rtcin() and writertc().
Move rtc initialization to cpu_initclocks().
Merge enablertclock() with cpu_initclocks() and remove enablertclock().
The extra entry point was just a leftover from 1.1.5.
timestamps for an atomic operation such as rename() on a local file
system to be identical.
Uniformize yet another idempotency ifdef. The comment nesting was
bogus.
of attributes so that `cp -p' to an msdos file system can succeed under
favourable circumstances (no uid or gid changes and no nonzero flags
except SF_ARCHIVED).
msdosfs_vnops.c:
The in-core inode flags were confused with the on-disk inode flags, so
chflags() clobbered the lock flag and caused a panic.
denode.h, msdosfs_denode.c, msdosfs_vnops.c:
Support the msdosfs archive attibute (ATTR_ARCHIVE) by mapping it to
the complement of the SF_ARCHIVED flag and setting the ATTR_ARCHIVE
bit when a file's modification time is set (but not when a file's
permissions are set; this is the standard wrong DOS behaviour).
denode.h, msdosfs_denode.c:
Remove the DE_UPDAT() macro. It was only used once, and the corresponding
macro in ufs has already been removed.
denode.h:
Don't change the timestamp for directories in DE_TIMES() (be consistent
with deupdat()).
msdosfs_vnops.c:
Handle chown() better: return EPERM instead of EINVAL if there are
insufficient permissions; otherwise, allow null changes.
for wd (they both count the number of sectors). The wpms stat is still
moderately bogus for all drivers. Even the count stat could be handled
better (partial blocks should be counted as full blocks; should errors
and retries be counted?).
Fix single-stepping of emulated FPU instructions.
Don't panic if an FPU instruction is attempted but there is no FPU
and no FPU emulator is configured.
boot(). This is needed so the "serialboot" stuff can share this file,
too.
Everything is #ifdef'ed so it evaluates to nothing when actually been
built in the "biosboot" directory.
The files in this directory are modified version of "biosboot". The
only difference is in that they perform their I/O via a serial port,
so their preferrable usage is to form bootblocks for systems where the
kernel happens to have an "options COMCONSOLE". Most of the code is
actually shared with "biosboot", and make will not (and should not)
descend into this directory by default. It is in the responsibility
of the user to build these bootblocks instead of the original ones.
short, it gets filled uop to its length. This matches the getdomainname
and gethostname manual pages.
(getbootfile also uses this function and I think it should have the same
behaviour)
This also fixes a bug with keyinit where the seed was not saved in
/etc/skeykeys. So S/Key should be fully functional again.
Reviewed by:
Submitted by:
Obtained from:
is supposed to work except "media removal" (shutting down the vn driver)
while some partitions are open.
Fix some errnos: return ENOTTY, not ENXIO for unknown ioctls; return
ENODEV, not ENXIO for the unsupported dump operation.
the APM-bios.
This stabilizes a couple of APM bioses quite a bit.
They all make the mistake of going into 16-bit mode, without clearing the
top half of the 32bit registers.
Later they do a
| movw %si,$0x7331
| movw %ax,0x6(%si)
or something along those lines and crash and burn, because their segment
is already relocated, so adding 0xf0171ce9 to the base of it is bad news.
At least SystemSoft is guilty of this bummer.
Voxware hackers should feel free to work on this some more, it's by no means
a perfect product.
(I have patches for GUS users running 2.x to run their GUS with bidirectional
DMA (talk while listening. All other soundboards must use push-to-talk until
people learn to build real hardware).
Submitted by: amancio hasty & paul traina
and one set by the protocol family. Also add another parameter to
rtalloc1() to allow for any interface flags to be ignored; currently
this is only useful for RTF_PRCLONING. Get rid of rt_prflags and re-unite
with rt_flags. Add T/TCP ``route metrics''.
NB: YOU MUST RECOMPILE `route' AND OTHER RELATED PROGRAMS AS A RESULT OF
THIS CHANGE.
This also adds a new interface parameter, `ifi_physical', which will
eventually replace IFF_ALTPHYS as the mechanism for specifying the
particular physical connection desired on a multiple-connection card.
NB: YOU MUST RECOMPILE `ifconfig' AND OTHER RELATED PROGRAMS AS A RESULT OF
THIS CHANGE.
DE_UPDATE was confused with DE_MODIFIED in some places (they do have
confusing names). Handle them exactly the same as IN_UPDATE and
IN_MODIFIED. This fixes chmod() and chown() clobbering the mtime
and other bugs.
DE_MODIFIED was set but not used.
Parenthesize macro args.
DE_TIMES() now takes a timeval arg instead of a timespec arg. It was
stupid to use a macro for speed and do unused conversions to prepare
for the macro.
Restore the left shifting of the DOS seconds count by 1. It got
lost among the shifts for the bitfields, so DOS seconds counts
appeared to range from 0 to 29 seconds (step 1) instead of 0 to 58
seconds (step 2).
Actually use the passed-in mtime in deupdat() as documented so that
utimes() works.
Change `extern __inline's to `static inline's so that msdosfs_fat.o
can be linked when it is compiled without -O.
Remove faking of directory mtimes to always be the current time. It's
more surprising for directory mtimes to change when you read the
directories than for them not to change when you write the directories.
This should be controlled by a mount-time option if at all.
Improve hzto():
Round up instead of down and then add 1 tick. This fixes sleep(1)
sometimes sleeping for < 1 second and usleep(10000) sometimes sleeping
for as little as 1 usec + syscall time.
Don't do all the calculations at splhigh().
Don't depend on `tick' being a multiple of 1000.
Don't lose accuracy for `sec' between 0x7fffffff / 1000 - 1000 and
0x7fffffff / hz.
Don't assume that longs are 32 bits or that ints have the same size as
longs.
Prepare for disk slices - more macros for handling disk device numbers,
version of readdisklabel() without DOS goop.
Clean up prototypes.
Uniformize idempotency ifdef.
diskslice.h:
New file.
dkbad.h:
Define more magic numbers.
Declare internalized version of dkbad struct and functions to operate on it.
Uniformize idempotency ifdef.
worked in the past only because of good fortune. Anyway, use the contig alloc
routine I wrote awhile ago (vm_page_alloc_contig) for the sound code to do
this allocation. Also, specify read+write on the permissions to pmap_enter().
Specifying just read can have unexpected consquences.
of returning EINVAL since something may depend on them being broken.
Allowing negative limits caused bugs almost everywhere. The recent
fixes for MAXSSIZ checked the limits too late to stop anyone defeating
limits set by root...
<string.h> isn't supposed to be used by the kernel.
cronix.h is <machine/cronix.h>, not "cronyx.h" (ambiguous) or
<sys/cronyx.h> (nonexistent; caused compile to fail).
cxreg.h is <i386/isa/cxreg.h>, not "cxreg.h".
<i386/isa/cpufunc.h> shouldn't be included directly; it is always
included by <sys/systm.h>.
<i386/include/*.h> is <machine/*.h>
<systm.h> is <sys/systm.h>.
<kernel.h> is <sys/kernel.h>.
<bpfilter.h> is "bpfilter.h". It really is in the current directory.
>Description:
If a process attempts to open a floppy tape device when the
device has been configured in the kernel, but did not probe and attach
on bootup, then a panic will occur.
[Review: The current ft situation is a crock, and this only bandaids
an earlier wound inflicted by making the attach conditional. This urgently
requires a review]
Submitted by: gene
Keep track of interrupt nesting level. It is normally 0
for syscalls and traps, but is fudged to 1 for their exit
processing in case they metamorphose into an interrupt
handler.
i386/genassym.c;
Remove support for the obsolete pcb_iml and pcb_cmap2.
Add support for pcb_inl.
i386/swtch.s:
Fudge the interrupt nesting level across context switches and in
the idle loop so that the work for preemptive context switches
gets counted as interrupt time, the work for voluntary context
switches gets counted mostly as system time (the part when
curproc == 0 gets counted as interrupt time), and only truly idle
time gets counted as idle time.
Remove obsolete support (commented out and otherwise) for pcb_iml.
Load curpcb just before curproc instead of just after so that
curpcb is always valid if curproc is. A few more changes like
this may fix tracing through context switches.
Remove obsolete function swtch_to_inactive().
include/cpu.h:
Use the new interrupt nesting level variable to implement a
non-fake CLF_INTR() so that accounting for the interrupt state
works.
You can use top, iostat or (best) an up to date systat to see
interrupt overheads. I see the expected huge interrupt overheads
for ISA devices (on a 486DX/33, about 55% for an IDE drive
transferring 1250K/sec and the same for a WD8013EBT network card
transferring 1100K/sec). The huge interrupt overheads for serial
devices are unfortunately normally invisible.
include/pcb.h:
Remove the obsolete pcb_iml and pcb_cmap2. Replace them by
padding to preserve binary compatibility.
Use part of the new padding for pcb_inl.
isa/icu.s:
isa/vector.s:
Keep track of interrupt nesting level.
used as an address value. Then all comparisons should be done unsigned
and not signed. Fix it with a typecast of u_quad_t.
Error can be demonstrated with the current bash in port, do a
ulimit -s unlimited and the machine hangs. bash delivers through
an internal error a large negative value for the stacksize, the
comparison saw this smaller than MAXSSIZ and then tried to expand
the stack to this size.
Now floppy tape support is *disabled* unless you specifically
request otherwise. Poul wanted it this way, and I guess I'm not going to argue
though it may seem counter-intuitive. We can always change it back, later.
make the sequencer code fully compatible with the aic7870 (ie 294x adaptors).
I've also added to my local mods putting the sequencer into "FASTMODE" clock.
This gives upwards of 2M/sec write preformance improvement in some scenarios.
There haven't been any reports of this causing problems, and I have been
reaping the benifits of it for more than a week now.
This also includes a new version of the pre-generated file <ugh>
Obtained from: John Aycock (aycock@cpsc.ucalgary.ca) and myself
flags & 0x1. Somebody should build a kernel with this and see if
the floppy-tape damaged people can turn it off properly with userconfig.
I can't reproduce the original problem here.
This should have been disabled for some time, but I had screwed up ...
This made spurious values appear for fd0 in systat, when there was
NCR SCSI activity.
comconsole will behave as expected. The true problem should be fixed
instead, Bruce' comment for this:
>Anyway, i found the reason for my problems: somehow, ICRNL isn't in
>effect at `userconfig' time (but only for comconsole?), hence only
ICRNL doesn't apply to cngetc(). cnputc() unconditionally does the
equivalent of ONLCR; perhaps cngetc() should unconditionally do the
equivalent of ICRNL. Ddb must be checking for CR. Userconfig only
checks for NL. Userconfig works with syscons because pccngetc()
does the conversion. This is probably the wrong place to do it.
Allow chown() to return success if the gid isn't changed even if
the gid is not the caller's. Such gids are normal for files created
in world-writable directories sucj as /tmp. This "fixes" annoying
error messages for mv'ing files created in /tmp to another file
system. mv still preserves the foreign gid of /tmp, but now does
it silently.
operation of each clist. Limit the growth of each clist. Clists
can only grow larger than the reserved minimum if there are free
cblocks in a shared pool. The size of this pool is now fixed
(this could be improved). The reserved and maximum sizes are more
carefully allocated for slip and ppp, depending on the mtu. A maximum
MTU of 16384 is now enforced for ppp.
1. The pageout daemon used to block under certain
circumstances, and we needed to add new functionality
that would cause the pageout daemon to block more often.
Now, the pageout daemon mostly just gets rid of pages
and kills processes when the system is out of swap.
The swapping, rss limiting and object cache trimming
have been folded into a new daemon called "vmdaemon".
This new daemon does things that need to be done for
the VM system, but can block. For example, if the
vmdaemon blocks for memory, the pageout daemon
can take care of it. If the pageout daemon had
blocked for memory, it was difficult to handle
the situation correctly (and in some cases, was
impossible).
2. The collapse problem has now been entirely fixed.
It now appears to be impossible to accumulate unnecessary
vm objects. The object collapsing now occurs when ref counts
drop to one (where it is more likely to be more simple anyway
because less pages would be out on disk.) The original
fixes were incomplete in that pathological circumstances
could still be contrived to cause uncontrolled growth
of swap. Also, the old code still, under steady state
conditions, used more swap space than necessary. When
using the new code, users will generally notice a
significant decrease in swap space usage, and theoretically,
the system should be leaving fewer unused pages around
competing for memory.
Submitted by: John Dyson
and into ether_input(). It was silly to have bpf want this one way and
ether_input want it another way. Ripped out trailer support from the few
remaining drivers that still had it.
1) make #includes correct
2) fix bugs in address check macros
3) fixed bugs in, and enabled, recopy if heavily fragmented code
4) moved call to bpf tap to be before enqueing packet (probably gratuitous)
5) fixed bug that caused "abnormal interrupt" at boot time/first use
6) added support for reading Zynx address ROM
7) fixed bug that caused broadcasts to not work shortly after booting (only
manifested if not using multicast - e.g. not in FreeBSD 2.0)
8) fixed spelling errors in comments
Submitted by: Matt Thomas
Subject: Mea culpa -- small fix for netboot fixes
In accordance with the unavoidable principle sof Murphy's Law, I discovered
that the fixes I recently contributed for the netboot code had some small
flaws in them. Two of them were just typos and had no effect on how the
program functioned. The other one was a missing line from the rootopts and
swapopts functions I created in bootmenu.c, which was supposed to initialize
the NFS sotype flag. It defaults to UDP, and you can change it to TCP with
the rootopts or swapopts commands, but then you can't change it back again.
I originally had a line at the top of each function to reinitialize this
flag, but somehow it got lost in the shuffle, probably because I don't
actually have a need for that flag yet.
Submitted by: wpaul
That was the good news. The bad news is that bad144 is a proper mess,
and I don't have time to fix it now, so you will probably not be able to
use it anyway.
Sorry guys, go out and buy a 100Mb IDE drive and a paddleboard :-(
If somebody wants to pick up on this: bad144 needs to learn how to
stay inside our slice of the disk. That's the trick.
