]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoPatches backported from later development version of OpenSSH which prevent
kris [Mon, 12 Feb 2001 06:44:51 +0000 (06:44 +0000)]
Patches backported from later development version of OpenSSH which prevent
(instead of just mitigating through connection limits) the Bleichenbacher
attack which can lead to guessing of the server key (not host key) by
regenerating it when an RSA failure is detected.

Reviewed by: rwatson

23 years agoFix `make depend' => `make kernel-depend'. Using the dependancy file
obrien [Mon, 12 Feb 2001 05:55:33 +0000 (05:55 +0000)]
Fix `make depend' => `make kernel-depend'.  Using the dependancy file
when rebuilding it is just Wrong.

23 years agoGive some additional DNS black hole possibilities as comments.
gshapiro [Mon, 12 Feb 2001 05:26:11 +0000 (05:26 +0000)]
Give some additional DNS black hole possibilities as comments.

Submitted by: clive

23 years agoIt sounded like a good idea at the time. The previous change breaks
peter [Mon, 12 Feb 2001 03:31:23 +0000 (03:31 +0000)]
It sounded like a good idea at the time.  The previous change breaks
FILE *buffer = stdout;
so back it out for now.

23 years agoChange all instances of `CURPROC' and `CURTHD' to `curproc,' in order
bmilekic [Mon, 12 Feb 2001 03:15:43 +0000 (03:15 +0000)]
Change all instances of `CURPROC' and `CURTHD' to `curproc,' in order
to stay consistent.

Requested by: bde

23 years agoCatch up to mutex API changes.
bmilekic [Mon, 12 Feb 2001 03:14:57 +0000 (03:14 +0000)]
Catch up to mutex API changes.

23 years agoTake advantage of the current libc sizeof(FILE) breakage (__sF[]) and
peter [Mon, 12 Feb 2001 02:50:30 +0000 (02:50 +0000)]
Take advantage of the current libc sizeof(FILE) breakage (__sF[]) and
try a hopefully more robust stdin/stdout/stderr.  This costs an indirect
pointer fetch, but saves us from changes in 'FILE'.  The __stdin stuff
is there to not pollute application name space if the application does
not use <stdio.h> and also in case something depended on the current
behavior where stdin etc was a #define.

Reviewed by: eischen, dillon

23 years agoo Fix build of libc broken in revision 1.2. offsetof() requires the
rwatson [Mon, 12 Feb 2001 02:44:33 +0000 (02:44 +0000)]
o Fix build of libc broken in revision 1.2.  offsetof() requires the
  inclusion of stddef.h.

Reviewed by: peter

23 years agoNote that packets read as outgoing have invalid IP header checksums.
archie [Mon, 12 Feb 2001 02:03:28 +0000 (02:03 +0000)]
Note that packets read as outgoing have invalid IP header checksums.

23 years agoCatch up to new priority interface.
jake [Mon, 12 Feb 2001 00:21:38 +0000 (00:21 +0000)]
Catch up to new priority interface.

23 years agoImplement a unified run queue and adjust priority levels accordingly.
jake [Mon, 12 Feb 2001 00:20:08 +0000 (00:20 +0000)]
Implement a unified run queue and adjust priority levels accordingly.

- All processes go into the same array of queues, with different
  scheduling classes using different portions of the array.  This
  allows user processes to have their priorities propogated up into
  interrupt thread range if need be.
- I chose 64 run queues as an arbitrary number that is greater than
  32.  We used to have 4 separate arrays of 32 queues each, so this
  may not be optimal.  The new run queue code was written with this
  in mind; changing the number of run queues only requires changing
  constants in runq.h and adjusting the priority levels.
- The new run queue code takes the run queue as a parameter.  This
  is intended to be used to create per-cpu run queues.  Implement
  wrappers for compatibility with the old interface which pass in
  the global run queue structure.
- Group the priority level, user priority, native priority (before
  propogation) and the scheduling class into a struct priority.
- Change any hard coded priority levels that I found to use
  symbolic constants (TTIPRI and TTOPRI).
- Remove the curpriority global variable and use that of curproc.
  This was used to detect when a process' priority had lowered and
  it should yield.  We now effectively yield on every interrupt.
- Activate propogate_priority().  It should now have the desired
  effect without needing to also propogate the scheduling class.
- Temporarily comment out the call to vm_page_zero_idle() in the
  idle loop.  It interfered with propogate_priority() because
  the idle process needed to do a non-blocking acquire of Giant
  and then other processes would try to propogate their priority
  onto it.  The idle process should not do anything except idle.
  vm_page_zero_idle() will return in the form of an idle priority
  kernel thread which is woken up at apprioriate times by the vm
  system.
- Update struct kinfo_proc to the new priority interface.  Deliberately
  change its size by adjusting the spare fields.  It remained the same
  size, but the layout has changed, so userland processes that use it
  would parse the data incorrectly.  The size constraint should really
  be changed to an arbitrary version number.  Also add a debug.sizeof
  sysctl node for struct kinfo_proc.

23 years ago1. The key SSD_KEY_RECOVERED_ERROR is not an error at all and should
mjacob [Sun, 11 Feb 2001 23:46:54 +0000 (23:46 +0000)]
1. The key SSD_KEY_RECOVERED_ERROR  is not an error at all and should
not be retried. It is an indication that there was an error that was
corrected during the execution of the command. This is per ANSI SCSI2
spec.

It's possible that these should also be noted to the console (as indicative,
perhaps, of growing media defect lists in drives), but the default of
printing errors out if bootverbose in this case is probably enough.

Also, there'd been a missing ERESTART for that clause anyway.

2. If you have an ABORTED COMMAND, it's almost invariably a SCSI parity
error. You should never be silent about these since users should do something
about this if it occurs (moving that power cord *away* from the SCSI cable is
always a good first start). This should print irrespective of bootverbose
because it's an actual real error even if we retry a transmission.

Reviewed by: audit@freebsd.org, gibbs@freebsd.org

23 years agoRemove (int) file descriptor locking. It should be up to the
deischen [Sun, 11 Feb 2001 22:07:32 +0000 (22:07 +0000)]
Remove (int) file descriptor locking.  It should be up to the
application to provide locking for I/O operations.  This doesn't
break any of my tests, but the old behavior can be restored by
compiling with _FDLOCKS_ENABLED.  This will eventually be removed
when it is obvious it does not cause any problems.

Remove most of flockfile implementation, with the exception of
flockfile_debug.

Make error messages more informational (submitted by Mike Heffner
<spock@techfour.net>, who's now known as mikeh@FreeBSD.org).

23 years agolibc MT-safety, part 2.
deischen [Sun, 11 Feb 2001 22:06:43 +0000 (22:06 +0000)]
libc MT-safety, part 2.

Add a lock to FILE.  flockfile and friends are now implemented
(for the most part) in libc.  flockfile_debug is implemented in
libc_r; I suppose it's about time to kill it but will do it in
a future commit.

Fix a potential deadlock in _fwalk in a threaded environment.
A file flag (__SIGN) was added to stdio.h that, when set, tells
_fwalk to ignore it in its walk.  This seemed to be needed in
refill.c because each file needs to be locked when flushing.

Add a stub for pthread_self in libc.  This is needed by flockfile
which is allowed by POSIX to be recursive.

Make fgetpos() error return value (-1) match man page.

Remove recursive calls to locked functions (stdio); I think I've
got them all, but I may have missed a couple.

A few K&R -> ANSI conversions along with removal of a few instances
of "register".

$Id$ -> $FreeBSD$ in libc/stdio/rget.c

Not objected to: -arch, a few months ago

23 years agolibc MT-safety, part 2.
deischen [Sun, 11 Feb 2001 22:04:18 +0000 (22:04 +0000)]
libc MT-safety, part 2.

Add a lock to FILE and define an additional flag.

23 years agoUname is in section 3 of the manual pages.
dwmalone [Sun, 11 Feb 2001 21:19:31 +0000 (21:19 +0000)]
Uname is in section 3 of the manual pages.

PR: 25008
Submitted by: fullermd@over-yonder.net

23 years agoMention PROT_NONE in the list of possible protections.
nik [Sun, 11 Feb 2001 19:30:41 +0000 (19:30 +0000)]
Mention PROT_NONE in the list of possible protections.

Pointed out by: kris

23 years agoInclude mmap(2) in the list of memory allocation functions.
nik [Sun, 11 Feb 2001 19:28:36 +0000 (19:28 +0000)]
Include mmap(2) in the list of memory allocation functions.

Reviewed by: hackers

23 years ago.Xr to mmap.
nik [Sun, 11 Feb 2001 18:53:50 +0000 (18:53 +0000)]
.Xr to mmap.

23 years agoNote that mmap(2) can allocate memory, as well as mapping existing files,
nik [Sun, 11 Feb 2001 18:51:17 +0000 (18:51 +0000)]
Note that mmap(2) can allocate memory, as well as mapping existing files,
in the .Nd.

Reviewed by: hackers

23 years agoLSI/SYMBIOS/NCR 53C[8XX|10XX] supported hardware updated.
groudier [Sun, 11 Feb 2001 18:13:15 +0000 (18:13 +0000)]
LSI/SYMBIOS/NCR 53C[8XX|10XX] supported hardware updated.
(One may check if the 8 words added are correct english :) )

23 years agoAdd a man page for the dbm_* functions, and update the Makefile to link
nik [Sun, 11 Feb 2001 17:24:25 +0000 (17:24 +0000)]
Add a man page for the dbm_* functions, and update the Makefile to link
it in.

Some review from -hackers (some time ago), and I think the best way to
get this improved (if it needs improving) or updating, is to bring it in.

PR:             docs/12557
Submitted by:   Tim Singletary <tsingle@triana.gsfc.nasa.gov>

23 years agoMake a big improvement to entropy-harvesting speed by not having any
markm [Sun, 11 Feb 2001 16:21:35 +0000 (16:21 +0000)]
Make a big improvement to entropy-harvesting speed by not having any
locks (only atomic assigns) in the harvest ringbuffer.

23 years agoAdd ja_JP.EUC locale support for LC_MESSAGES, LC_MONETARY, and LC_NUMERIC.
knu [Sun, 11 Feb 2001 16:19:43 +0000 (16:19 +0000)]
Add ja_JP.EUC locale support for LC_MESSAGES, LC_MONETARY, and LC_NUMERIC.

23 years agoAdd ko_KR.EUC locale support for LC_MESSAGES, LC_MONETARY, and LC_NUMERIC.
asmodai [Sun, 11 Feb 2001 15:40:01 +0000 (15:40 +0000)]
Add ko_KR.EUC locale support for LC_MESSAGES, LC_MONETARY, and LC_NUMERIC.

Submitted by: CHOI Junho <cjh@kr.FreeBSD.org>

23 years agoFix:
groudier [Sun, 11 Feb 2001 15:38:06 +0000 (15:38 +0000)]
Fix:
- Missing cpu_to_scr() added (endian-ness).

Improvement (fix|workaroung??):
- Blindly firing a PPR can lead to some messy situations due to
  various causes or misfeatures, for example:
  * The 53C1010-[33|66] supports offset 62 in DT mode, but only
    offset 31 in ST mode. As a result, a PPR(DT, offset 62)
    responded with PPR(ST, any offset > 31) must be rejected.
  * A device that doesn't know about PPR should reject it, but
    may also be confused by this message.
  When a PPR encounters problems, the driver now patches the goal
  transfer settings for legacy negotiations to be performed later
  with the offending target. This give a chance for bad situations
  to be fixed automagically.

23 years agoDon't use hardcoded struct size, use offsetof() instead (make size calculations
phantom [Sun, 11 Feb 2001 15:09:31 +0000 (15:09 +0000)]
Don't use hardcoded struct size, use offsetof() instead (make size calculations
dynamic)

23 years agoDon't try to convert grouping strings in case if C or POSIX locale
phantom [Sun, 11 Feb 2001 15:07:26 +0000 (15:07 +0000)]
Don't try to convert grouping strings in case if C or POSIX locale
was explicitly specified.

Submitted by: ache

23 years agoRIP <machine/lock.h>.
markm [Sun, 11 Feb 2001 10:44:09 +0000 (10:44 +0000)]
RIP <machine/lock.h>.

Some things needed bits of <i386/include/lock.h> - cy.c now has its
own (only) copy of the COM_(UN)LOCK() macros, and IMASK_(UN)LOCK()
has been moved to <i386/include/apic.h> (AKA <machine/apic.h>).
Reviewed by: jhb

23 years agoClean up RST ratelimiting. Previously, ratelimiting occured before tests
bmilekic [Sun, 11 Feb 2001 07:39:51 +0000 (07:39 +0000)]
Clean up RST ratelimiting. Previously, ratelimiting occured before tests
were performed to determine if the received packet should be reset. This
created erroneous ratelimiting and false alarms in some cases. The code
has now been reorganized so that the checks for validity come before
the call to badport_bandlim. Additionally, a few changes in the symbolic
names of the bandlim types have been made, as well as a clarification of
exactly which type each RST case falls under.

Submitted by: Mike Silbersack <silby@silby.com>

23 years agoLong awaited style fixup in mbuf code. Get rid of K&R style prototyping
bmilekic [Sun, 11 Feb 2001 05:02:06 +0000 (05:02 +0000)]
Long awaited style fixup in mbuf code. Get rid of K&R style prototyping
and function argument declarations. Make sure that functions that are
supposed to return a pointer return NULL in case of failure. Don't cast
NULL. Finally, get rid of annoying `register' uses.

23 years agoEliminate ISP2100_FABRIC- we always allow for fabric now. Add an
mjacob [Sun, 11 Feb 2001 03:56:48 +0000 (03:56 +0000)]
Eliminate ISP2100_FABRIC- we always allow for fabric now. Add an
isp_iid_set/isp_iid for fibre channel- this is because we now
fake a port database entry for ourselves. Add the additional loop
states between LOOP_PDB_RCVD and LOOP_READY.

Change and comment on a wad of Fibre Channel isp_control functions.
Change and comment on some of the ISPASYNC Fibre Channel events.

23 years agoAdd structure defining FC-AL position maps. The only tool that I know of
mjacob [Sun, 11 Feb 2001 03:53:58 +0000 (03:53 +0000)]
Add structure defining FC-AL position maps. The only tool that I know of
that really uses this is luxadm(8) under Solaris.

23 years agoShuffle around how we do isp_disable management- make sure we return 0 so
mjacob [Sun, 11 Feb 2001 03:53:23 +0000 (03:53 +0000)]
Shuffle around how we do isp_disable management- make sure we return 0 so
the unit number doesn't get reused.

Make sure that if we've compiled for ISP_TARGET_MODE we set the
default role to be ISP_ROLE_INITIATOR|ISP_ROLE_TARGET.

Do some misc other cleanups.

23 years agoAdd isp_fc_runstate function- this function's purpose is to, in stages,
mjacob [Sun, 11 Feb 2001 03:52:04 +0000 (03:52 +0000)]
Add isp_fc_runstate function- this function's purpose is to, in stages,
and depending on role, make sure link is up, scan the fabric (if we're
connected to a fabric), scan the local loop (if appropriate), merge
the results into the local port database then, check once again
to make sure we have f/w at FW_READY state and the the loopstate
is LOOP_READY.

23 years agoDefault the plist dir as we know where it is relative to us.
obrien [Sun, 11 Feb 2001 03:50:32 +0000 (03:50 +0000)]
Default the plist dir as we know where it is relative to us.

23 years agoMinor comment & doc tweak.
obrien [Sun, 11 Feb 2001 03:49:34 +0000 (03:49 +0000)]
Minor comment & doc tweak.

23 years agoRoll minor version. Remove ISP2100_FABRIC define (unneeded now).
mjacob [Sun, 11 Feb 2001 03:48:54 +0000 (03:48 +0000)]
Roll minor version. Remove ISP2100_FABRIC define (unneeded now).

Comment out usage of ISP_SMPLOCK- I have my doubts that this works sanely
as yet because CAM itself still needs Giant. I *was* dropping my lock
and grabbing Giant when doing the upcall for completion, but this is all
seems ridiculous until CAM is fixed.

23 years agoDo some cleanup based upon adapter role- mainly not enabling interrupts
mjacob [Sun, 11 Feb 2001 03:47:39 +0000 (03:47 +0000)]
Do some cleanup based upon adapter role- mainly not enabling interrupts
if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE.
Make sure we note if something is a fabric device.

Target mode:
Finally fix (to a first approximation) SCSI Target Mode again- we needed
to correctly check against CAM_TARGET_WILDCARD and CAM_LUN_WILDCARD
so that targbh won't confuse us. Comment out the drainqueue stuff for
now. Use isp_fc_runstate instead if isp_control/ISPCTL_FCLINK_TEST.

23 years agoMinor stuff:
mjacob [Sun, 11 Feb 2001 03:44:43 +0000 (03:44 +0000)]
Minor stuff:

Remove ISP2100_FABRIC defines- we always handle fabric now. Insert
isp_getmap helper function (for getting Loop Position map). Make
sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE
for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic
f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP,
and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications
upstream for changes in the name server, changes in the port database, and
f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event.

Major stuff:
Quite massively redo how we handle Loop events- we've now added several
intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us
a lot finer control about how we scan fabric, whether we go further
than scanning fabric, how we look at the local loop, and whether we
merge entries at the level or not. This is the next to last step for
moving managing loop state out of the core module entirely (whereupon
loop && fabric events will simply freeze the command queue and a thread
will run to figure out what's changed and *it* will re-enable the queu).
This fine amount of control also gets us closer to having an external
policy engine decide which fabric devices we really want to log into.

23 years agoupdate to latest and greatest f/w
mjacob [Sun, 11 Feb 2001 03:35:45 +0000 (03:35 +0000)]
update to latest and greatest f/w

23 years ago- Place back STR string declarations for lock/unlock strings used for KTR_LOCK
bmilekic [Sun, 11 Feb 2001 02:54:16 +0000 (02:54 +0000)]
- Place back STR string declarations for lock/unlock strings used for KTR_LOCK
  tracing in order to avoid duplication.
- Insert some tracepoints back into the mutex acq/rel code, thus ensuring
  that we can trace all lock acq/rel's again.
- All CURPROC != NULL checks are MPASS()es (under MUTEX_DEBUG) because they
  signify a serious mutex corruption.
- Change up some KASSERT()s to MPASS()es, and vice-versa, depending on the
  type of problem we're debugging (INVARIANTS is used here to check that
  the API is being used properly whereas MUTEX_DEBUG is used to ensure that
  something general isn't happening that will have bad impact on mutex
  locks).

Reminded by: jhb, jake, asmodai

23 years agoUse decimal point from locale
ache [Sun, 11 Feb 2001 02:25:56 +0000 (02:25 +0000)]
Use decimal point from locale

23 years agoChange localizing to LC_ALL
ache [Sat, 10 Feb 2001 23:25:09 +0000 (23:25 +0000)]
Change localizing to LC_ALL

23 years agoTemporary workaround to get things to compile. I could have updated
mjacob [Sat, 10 Feb 2001 23:22:49 +0000 (23:22 +0000)]
Temporary workaround to get things to compile. I could have updated
genassym here, but what I've also noticed is that we're dorking
with a mutex directly at assembler level- I'm not sure that this
is wise at this stage in the SMP port- I think it's going to be much
safer for a while to do things in C until SMP wunderkind figure out
what works and slow down this 3 order differential...

23 years agoAdd missing 4-th space at the end of int_curr_symbol
ache [Sat, 10 Feb 2001 23:07:05 +0000 (23:07 +0000)]
Add missing 4-th space at the end of int_curr_symbol

23 years agoResurrect Minix sh(1), after its license has been changed to BSD one.
abial [Sat, 10 Feb 2001 22:57:33 +0000 (22:57 +0000)]
Resurrect Minix sh(1), after its license has been changed to BSD one.

23 years agoLocalize it (LC_NUMERIC)
ache [Sat, 10 Feb 2001 22:46:47 +0000 (22:46 +0000)]
Localize it (LC_NUMERIC)

23 years agoAdd Polish locale.
abial [Sat, 10 Feb 2001 22:44:31 +0000 (22:44 +0000)]
Add Polish locale.

23 years agoUnbreak by syncing with changes in <sys/sysctl.h> macros.
abial [Sat, 10 Feb 2001 22:36:05 +0000 (22:36 +0000)]
Unbreak by syncing with changes in <sys/sysctl.h> macros.

23 years agoClear the reschedule flag after finding it set in userret(). This
jake [Sat, 10 Feb 2001 20:33:35 +0000 (20:33 +0000)]
Clear the reschedule flag after finding it set in userret().  This
used to be in cpu_switch(), but I don't see any difference between
doing it here.

23 years agomake it possible to specify grouping number from range 0..CHAR_MAX,
phantom [Sat, 10 Feb 2001 20:22:45 +0000 (20:22 +0000)]
make it possible to specify grouping number from range 0..CHAR_MAX,
not only one-digit number

23 years ago-Document DS20 serial/graphics console (as it bit me yesterday.. again.. grr)
wilko [Sat, 10 Feb 2001 20:22:06 +0000 (20:22 +0000)]
-Document DS20 serial/graphics console (as it bit me yesterday.. again.. grr)
-CS20 can have 2 CPUs but we don't have SMP currently on alpha

23 years agoUse "namespace.h" and "un-namespace.h"
phantom [Sat, 10 Feb 2001 19:57:26 +0000 (19:57 +0000)]
Use "namespace.h" and "un-namespace.h"

Requested by: deischen

23 years agoAcquire sched_lock around need_resched() in roundrobin() to satisfy
jake [Sat, 10 Feb 2001 19:07:32 +0000 (19:07 +0000)]
Acquire sched_lock around need_resched() in roundrobin() to satisfy
assertions that it is held.  Since roundrobin() is a timeout there's
no possible way that it could be called with sched_lock held.

23 years agoBack out my "3;3" -> something change. While technikally the same, it only
ache [Sat, 10 Feb 2001 19:06:32 +0000 (19:06 +0000)]
Back out my "3;3" -> something change. While technikally the same, it only
leads to confusion.

el_GR: change "0;0" to "-1".
"0" is repeater for previous number, not no grouping

23 years agoBack out my "3;3" -> something changes. While technikally the same, it only
ache [Sat, 10 Feb 2001 18:51:51 +0000 (18:51 +0000)]
Back out my "3;3" -> something changes. While technikally the same, it only
leads to confusion.

23 years agoaic7xxx.c:
gibbs [Sat, 10 Feb 2001 18:04:27 +0000 (18:04 +0000)]
aic7xxx.c:
Style nits.

Make sure that our selection hardware is disabled
as soon as possible after detecting a busfree and
even go so far as to disable the selection hardware
in advance of an event that will cause a busfree
(ABORT or BUS DEVICE RESET message).  The concern
is that the selection hardware will select a target
for which, after processing the bus free, there
will be no commands pending.  The sequencer idle
loop will re-enable the selection should it still be
necessary.

In ahc_handle_scsiint(), clear SSTAT0 events several
PCI transactions (most notably reads) prior to clearing
SCSIINT.  The newer chips seem to take a bit of time to
see the change which can make the clearing of SCSIINT
ineffective.

Don't bother panicing at the end of ahc_handle_scsiint().
Getting to the final else just means we lost the race
with clearing SCSIINT.

In ahc_free(), handle init-level 0.  This can happen when we
fail the attach for RAID devices.  While I'm here, also kill
the parent dma tag.

In ahc_match_scb(), consider initiator ccbs to be any
that are not from the target mode group.  This fixes
a bug where an external target reset CCB was not getting
cleaned up by the reset code.

Don't bother freezing a ccb in any of our "abort" routines
when the status is set to CAM_REQ_CMP.  This can happen
for a target reset ccb.

aic7xxx.reg:
Reserve space for a completion queue.  This will be used
to enhance performance in the near future.

aic7xxx.seq:
Remove an optimization for the 7890 autoflush bug that
turned out to allow, in rare cases, some data to get
lost.

Implement a simpler, faster, fix for the PCI_2_1 retry
bug that hangs the sequencer on an SCB dma for certain chips.

Test against SAVED_SCSIID rather than SELID during target
reselections.  This is how we always did it in the past,
but the code was modified while trying to work around an
issue with the 7895.  SAVED_SCSIID takes into account
twin channel adapters such as the 2742T, whereas SELID
does not have the channel bit.  This caused invalid
selection warnings and other strangeness on these cards.

aic7xxx_pci.c
Use the correct mask for checking the generic aic7892
entry.

23 years agoA better fix for the PacketAliasProxyRule() call.
brian [Sat, 10 Feb 2001 17:26:14 +0000 (17:26 +0000)]
A better fix for the PacketAliasProxyRule() call.

Submitted by: Ian Dowse <iedowse@maths.tcd.ie>

23 years ago. Fix semantics of grouping (LC_MONETARY::mon_grouping,
phantom [Sat, 10 Feb 2001 15:36:46 +0000 (15:36 +0000)]
. Fix semantics of grouping (LC_MONETARY::mon_grouping,
  LC_NUMERIC::grouping) values.
. Always set __XXX_changed flags then loading numeric & monetary locale
  categories to allow localeconv() to use C locale also.

23 years agoAdd en_NZ.ISO_8859-1.
asmodai [Sat, 10 Feb 2001 13:46:59 +0000 (13:46 +0000)]
Add en_NZ.ISO_8859-1.

23 years agoactivate cs_CZ.ISO_8859-2 and el_GR.ISO_8859-7
phantom [Sat, 10 Feb 2001 13:35:30 +0000 (13:35 +0000)]
activate cs_CZ.ISO_8859-2 and el_GR.ISO_8859-7

23 years agoadd LC_NUMERIC, LC_MESSAGES and LC_MONETARY for cs_CZ.ISO_8859-2
phantom [Sat, 10 Feb 2001 13:31:36 +0000 (13:31 +0000)]
add LC_NUMERIC, LC_MESSAGES and LC_MONETARY for cs_CZ.ISO_8859-2

PR: misc/24970
Submitted by: Rudolf Cejka <cejkar@dcse.fee.vutbr.cz>

23 years agoAdd LC_NUMERIC, LC_MESSAGES and LC_MONETARY for el_GR.ISO_8859-7
phantom [Sat, 10 Feb 2001 13:28:23 +0000 (13:28 +0000)]
Add LC_NUMERIC, LC_MESSAGES and LC_MONETARY for el_GR.ISO_8859-7

Submitted by: Panagiotis Astithas <past@netmode.ntua.gr> via -i18n

23 years agoUse ${MACHINE_ARCH} instead of ${MACHINE} to support pc98.
nyan [Sat, 10 Feb 2001 13:11:02 +0000 (13:11 +0000)]
Use ${MACHINE_ARCH} instead of ${MACHINE} to support pc98.

23 years agoAdd pc98 support.
nyan [Sat, 10 Feb 2001 12:52:56 +0000 (12:52 +0000)]
Add pc98 support.

23 years agomdoc(7) police: polishing.
ru [Sat, 10 Feb 2001 10:51:39 +0000 (10:51 +0000)]
mdoc(7) police: polishing.

23 years agomdoc(7) police: mark LC_NUMERIC with .Dv.
ru [Sat, 10 Feb 2001 10:26:52 +0000 (10:26 +0000)]
mdoc(7) police: mark LC_NUMERIC with .Dv.

23 years agoRemove spurious '"'.
phk [Sat, 10 Feb 2001 08:30:21 +0000 (08:30 +0000)]
Remove spurious '"'.

PR: 24979
Submitted by: Joseph Scott <joseph@randomnetworks.com>

23 years agoImprove language and code examples.
nik [Sat, 10 Feb 2001 07:48:01 +0000 (07:48 +0000)]
Improve language and code examples.

PR:             docs/24961
Submmitted by:  Tony Finch <dot@dotat.at>

23 years agoFix nasty bug where make(1) assumed that you could read the directory it
will [Sat, 10 Feb 2001 07:12:18 +0000 (07:12 +0000)]
Fix nasty bug where make(1) assumed that you could read the directory it
was in.  This shall be MFC'd in about three days (probably not a good idea
to MFC the stylistic changes though - see below).

PR: 19978
Submitted by: Gerald Pfeifer <pfeifer@dbai.tuwien.ac.at>
Patch by: roam (slightly modified by me to use NULL not NIL)

23 years agoThere are now compat libs for the Alpha too.
obrien [Sat, 10 Feb 2001 07:07:09 +0000 (07:07 +0000)]
There are now compat libs for the Alpha too.

23 years agoThe compat[34]x dists need to have their machine arch appended.
obrien [Sat, 10 Feb 2001 07:06:40 +0000 (07:06 +0000)]
The compat[34]x dists need to have their machine arch appended.

23 years agoAdd 4x compat libs for Alpha.
obrien [Sat, 10 Feb 2001 07:00:26 +0000 (07:00 +0000)]
Add 4x compat libs for Alpha.
These come from the live FS of the BSDi FreeBSD/AXP 4.2 disc set.

Forgotten by: non-Alpha person

23 years agoLook to the future and make this easier to add libs to.
obrien [Sat, 10 Feb 2001 06:53:31 +0000 (06:53 +0000)]
Look to the future and make this easier to add libs to.

23 years agoLocalize it (LC_NUMERIC)
ache [Sat, 10 Feb 2001 06:49:53 +0000 (06:49 +0000)]
Localize it (LC_NUMERIC)

23 years agoReformat the lib list.
obrien [Sat, 10 Feb 2001 06:44:18 +0000 (06:44 +0000)]
Reformat the lib list.

23 years agoNote that decimal point taken from locale
ache [Sat, 10 Feb 2001 06:42:32 +0000 (06:42 +0000)]
Note that decimal point taken from locale

23 years agoAdd more 3.5.1 libs that don't exist in 5-CURRENT.
obrien [Sat, 10 Feb 2001 06:41:59 +0000 (06:41 +0000)]
Add more 3.5.1 libs that don't exist in 5-CURRENT.
These are: libalias.so.3 libfetch.so.1 libperl.so.3 libss.so.2 libvgl.so.1

23 years agoNote that decimal point taken from locale (SUSv2)
ache [Sat, 10 Feb 2001 06:32:03 +0000 (06:32 +0000)]
Note that decimal point taken from locale (SUSv2)

23 years agoReformat the lib list.
obrien [Sat, 10 Feb 2001 06:32:02 +0000 (06:32 +0000)]
Reformat the lib list.

23 years agolibc_r.so was bumped between 3.4 and 3.5[.1], so we need libc_r.so.4 also.
obrien [Sat, 10 Feb 2001 06:29:53 +0000 (06:29 +0000)]
libc_r.so was bumped between 3.4 and 3.5[.1], so we need libc_r.so.4 also.

23 years agoUpdate the i386 compat3x dist to the FreeBSD 3.5.1 libs.
obrien [Sat, 10 Feb 2001 06:25:45 +0000 (06:25 +0000)]
Update the i386 compat3x dist to the FreeBSD 3.5.1 libs.

I had moved on to FreeBSD 4.0 before 3.5-R and had forgotten to keep this
collection up to date.

23 years agoTake decimal point from locale instead of hardcoded '.' (SUSv2)
ache [Sat, 10 Feb 2001 06:25:33 +0000 (06:25 +0000)]
Take decimal point from locale instead of hardcoded '.' (SUSv2)

23 years agoNote that decimal point character taken from locale (SUSv2)
ache [Sat, 10 Feb 2001 05:52:59 +0000 (05:52 +0000)]
Note that decimal point character taken from locale (SUSv2)

23 years agoUse decimal point from localeconv() instead of hardcoded '.' (SUSv2)
ache [Sat, 10 Feb 2001 05:46:05 +0000 (05:46 +0000)]
Use decimal point from localeconv() instead of hardcoded '.' (SUSv2)

23 years agoNote the fact that decimal point taken from locale (according to SUSv2)
ache [Sat, 10 Feb 2001 05:16:41 +0000 (05:16 +0000)]
Note the fact that decimal point taken from locale (according to SUSv2)

23 years agoUse decimap_point from localeconv() instead of hardcoded '.'
ache [Sat, 10 Feb 2001 05:05:09 +0000 (05:05 +0000)]
Use decimap_point from localeconv() instead of hardcoded '.'

Obtained from: inspired by NetBSD strtod

23 years agoNote that crypto/ is not used to build in, people should see secure/
kris [Sat, 10 Feb 2001 04:47:47 +0000 (04:47 +0000)]
Note that crypto/ is not used to build in, people should see secure/
instead.

23 years agoAs temporary workaround for missing *grouping fields parser always return "no
ache [Sat, 10 Feb 2001 04:32:55 +0000 (04:32 +0000)]
As temporary workaround for missing *grouping fields parser always return "no
grouping" (CHAR_MAX, '\0').
Fixme: grouping parser needs to be implemented.

23 years agoAccording to Garrett, POSIX widely use -1 to indicate CHAR_MAX, so back out
ache [Sat, 10 Feb 2001 03:31:23 +0000 (03:31 +0000)]
According to Garrett, POSIX widely use -1 to indicate CHAR_MAX, so back out
all my "-1" -> "something" fixes and replace -1 with CHAR_MAX directly in
strtol() in cnv()

23 years agoUse -1 for CHAR_MAX instead of empty string as recommended by POSIX
ache [Sat, 10 Feb 2001 03:20:27 +0000 (03:20 +0000)]
Use -1 for CHAR_MAX instead of empty string as recommended by POSIX
Still not parsed in localeconv() in anycase

23 years agoI change my mind a bit: assuming last \0 is contrproductive since will cause
ache [Sat, 10 Feb 2001 02:55:42 +0000 (02:55 +0000)]
I change my mind a bit: assuming last \0 is contrproductive since will cause
user confusion, so specify it directly, i.e.  change "3" to "3;0".

In this style "3;" or "3" must not cause repeating
(converted to \3, CHAR_MAX, \0)
Still not implemented and broken in localeconv()

23 years agoI change my mind a bit: assuming last \0 is contrproductive since will cause
ache [Sat, 10 Feb 2001 02:51:20 +0000 (02:51 +0000)]
I change my mind a bit: assuming last \0 is contrproductive since will cause
user confusion, so specify it directly, i.e. change "3" to "3;0".
In this style "3;" must not cause repeating (converted to \3, CHAR_MAX, \0)
NOTE: still no proper conversion done in localeconv()

23 years agoReenable preemption on interrupts. My last commit accidentally reverted
jhb [Sat, 10 Feb 2001 02:46:50 +0000 (02:46 +0000)]
Reenable preemption on interrupts.  My last commit accidentally reverted
it as I was playing with some other ways of doing kernel preemption.

You must still specify the PREEMPTION option in your config file to get a
preemptive kernel.

23 years agoRe-enable preemption on interrupts. My last commit accidentally reverted
jhb [Sat, 10 Feb 2001 02:41:50 +0000 (02:41 +0000)]
Re-enable preemption on interrupts.  My last commit accidentally reverted
it as I was playing with some other ways of doing kernel preemption.

23 years agoReplace "3;3" with "3". This is cosmetique, all integer before \0 (i.e. \3)
ache [Sat, 10 Feb 2001 02:31:14 +0000 (02:31 +0000)]
Replace "3;3" with "3". This is cosmetique, all integer before \0 (i.e. \3)
repeated forever according to SUSv2

Remove "0;0" - \0 means not "no grouping" but repeat forever previous char,
and added automatically. Empty string could be parsed later into CHAR_MAX
(real "no grouping") by localeconv()

23 years ago- Make astpending and need_resched process attributes rather than CPU
jhb [Sat, 10 Feb 2001 02:20:34 +0000 (02:20 +0000)]
- Make astpending and need_resched process attributes rather than CPU
  attributes.  This is needed for AST's to be properly posted in a preemptive
  kernel.  They are backed by two new flags in p_sflag: PS_ASTPENDING and
  PS_NEEDRESCHED.  They are still accesssed by their old macros:
  aston(), astoff(), etc.  For completeness, an astpending() macro has been
  added to check for a pending AST, and clear_resched() has been added to
  clear need_resched().
- Rename syscall2() on the x86 back to syscall() to be consistent with
  other architectures.

23 years agoChange "3;3" in mon_grouping to "3". This is cosmetique, last \0 will cause
ache [Sat, 10 Feb 2001 02:17:49 +0000 (02:17 +0000)]
Change "3;3" in mon_grouping to "3". This is cosmetique, last \0 will cause
previous \3 repeat forever, according to SUSv2
Add comment about mon_grouping separator
NOTE: mon_grouping localeconv() parsing still broken.

23 years agoAdd a macro mtx_intr_enable() to alter a spin lock such that interrupts
jhb [Sat, 10 Feb 2001 02:15:18 +0000 (02:15 +0000)]
Add a macro mtx_intr_enable() to alter a spin lock such that interrupts
will be enabled when it is released.