]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoMake it clear that -P is the default.
ru [Fri, 25 May 2001 07:32:58 +0000 (07:32 +0000)]
Make it clear that -P is the default.

PR: docs/27629

23 years agoAdd intrack field to each slot. This can be used to acknowledge
imp [Fri, 25 May 2001 05:25:43 +0000 (05:25 +0000)]
Add intrack field to each slot.  This can be used to acknowledge
interrupts on other buses.  Right now it isn't used, but will be for
the pci attachment.

# Add copyright by me for this year since I've changed so much.

23 years agoMinor name space issues.
imp [Fri, 25 May 2001 05:22:00 +0000 (05:22 +0000)]
Minor name space issues.

23 years agoRFC 1323 (TCP extensions) is now on by default.
obrien [Fri, 25 May 2001 04:27:48 +0000 (04:27 +0000)]
RFC 1323 (TCP extensions) is now on by default.

23 years agoAdd support for
imp [Fri, 25 May 2001 02:24:15 +0000 (02:24 +0000)]
Add support for
card "D" "Link DWL-650 11Mbps WLAN Card"
which is the most amusing CIS mistake I've seen in some time.

# I'm using this card to make this commit!

23 years agoI didn't fix the comment in rev 1.107.
obrien [Fri, 25 May 2001 01:46:39 +0000 (01:46 +0000)]
I didn't fix the comment in rev 1.107.

23 years agoTurn on TCP_EXTENSIONS (rfc1323) by defualt.
obrien [Fri, 25 May 2001 01:43:42 +0000 (01:43 +0000)]
Turn on TCP_EXTENSIONS (rfc1323) by defualt.

23 years agoDon't rely on cdevsw_add() when we hack about with dev_t's.
phk [Thu, 24 May 2001 20:28:06 +0000 (20:28 +0000)]
Don't rely on cdevsw_add() when we hack about with dev_t's.

23 years agoDon't take the detour around devsw() to find out if the proto-cdevsw
phk [Thu, 24 May 2001 20:27:16 +0000 (20:27 +0000)]
Don't take the detour around devsw() to find out if the proto-cdevsw
is already initialized.

23 years agoBack out rev 1.4 (wint_t and mbstate_t) as I am tired of the complaints.
obrien [Thu, 24 May 2001 19:29:15 +0000 (19:29 +0000)]
Back out rev 1.4 (wint_t and mbstate_t) as I am tired of the complaints.
The STLport will probably become broken again, but I'll work on fixing it
later.

I wish someone would explain why the NetBSD Cirtus branch has the types
in their stddef.h...

Requested by: bde, ru
PR: 27606
Submitted by: Naohiko Tsuji <yakisoba@f2.dion.ne.jp>

23 years agoNetBSD's inetd uses different syntex from FreeBSD's in faith support.
ume [Thu, 24 May 2001 19:00:22 +0000 (19:00 +0000)]
NetBSD's inetd uses different syntex from FreeBSD's in faith support.

23 years agoAdd the "auth_as_self" option to the pam_unix module (there is no
markm [Thu, 24 May 2001 18:35:52 +0000 (18:35 +0000)]
Add the "auth_as_self" option to the pam_unix module (there is no
reason not to add it to others later). This causes the pam_unix
module to check the user's _own_ password, not the password of the
account that the user is authenticating into. This will allow eg:
WHEELSU type behaviour from su(1).

23 years agoOops. Unremove vestiges of the old, broken sound drivers. They are now
bde [Thu, 24 May 2001 18:19:00 +0000 (18:19 +0000)]
Oops.  Unremove vestiges of the old, broken sound drivers.  They are now
used in new drivers.

23 years agowhitespace/style
alfred [Thu, 24 May 2001 18:06:22 +0000 (18:06 +0000)]
whitespace/style

23 years agoStick VM syscalls back under Giant if the BLEED option is not defined.
jhb [Thu, 24 May 2001 18:04:29 +0000 (18:04 +0000)]
Stick VM syscalls back under Giant if the BLEED option is not defined.

23 years agoAdd a new kernel option 'BLEED' to be used for code that is still under
jhb [Thu, 24 May 2001 17:01:51 +0000 (17:01 +0000)]
Add a new kernel option 'BLEED' to be used for code that is still under
development but is being developed in the tree for whatever reason.

Not objected to by:     peter, jlemon

23 years agoRemove vestiges of the old, broken sound drivers.
bde [Thu, 24 May 2001 16:52:13 +0000 (16:52 +0000)]
Remove vestiges of the old, broken sound drivers.

23 years agoActually rename FDESC, PORTAL, UMAP and UNION file systems.
ru [Thu, 24 May 2001 15:20:11 +0000 (15:20 +0000)]
Actually rename FDESC, PORTAL, UMAP and UNION file systems.

OK'ed by: bp

23 years agomount_fdesc -> mount_fdescfs
ru [Thu, 24 May 2001 13:23:54 +0000 (13:23 +0000)]
mount_fdesc -> mount_fdescfs
mount_portal -> mount_portalfs

23 years agomount_umap(8) -> mount_umapfs(8).
ru [Thu, 24 May 2001 13:20:41 +0000 (13:20 +0000)]
mount_umap(8) -> mount_umapfs(8).

23 years agomount_null(8) -> mount_nullfs(8).
ru [Thu, 24 May 2001 13:17:47 +0000 (13:17 +0000)]
mount_null(8) -> mount_nullfs(8).

23 years agomount_portal -> mount_portalfs.
ru [Thu, 24 May 2001 13:15:49 +0000 (13:15 +0000)]
mount_portal -> mount_portalfs.

23 years agoMissed one mount_null(8) -> mount_nullfs(8) in previous commit.
ru [Thu, 24 May 2001 13:13:56 +0000 (13:13 +0000)]
Missed one mount_null(8) -> mount_nullfs(8) in previous commit.

23 years ago(Re-)enabled boot2.
nyan [Thu, 24 May 2001 11:19:43 +0000 (11:19 +0000)]
(Re-)enabled boot2.

23 years agoUpdate boot[12] program to compile ELF binary.
nyan [Thu, 24 May 2001 11:18:32 +0000 (11:18 +0000)]
Update boot[12] program to compile ELF binary.

Submitted by: Nobuyuki Koganemaru <n-kogane@syd.odn.ne.jp> and
kawanobe@st.rim.or.jp (Kawanobe Koh)

23 years agoRestrict even further what parts of <sys/conf.h> can be seen from
phk [Thu, 24 May 2001 09:33:08 +0000 (09:33 +0000)]
Restrict even further what parts of <sys/conf.h> can be seen from
userland.

23 years agoRemove the "undocumented" comment in re lflag.
nik [Thu, 24 May 2001 09:28:59 +0000 (09:28 +0000)]
Remove the "undocumented" comment in re lflag.

23 years agoDocument the -l option.
nik [Thu, 24 May 2001 09:28:05 +0000 (09:28 +0000)]
Document the -l option.

PR: docs/27440
Submitted by: Flix-Antoine Paradis <reel@sympatico.ca>

23 years agoIf the user exists abruptly, tip's "tipout" child can hang around
phk [Thu, 24 May 2001 09:27:02 +0000 (09:27 +0000)]
If the user exists abruptly, tip's "tipout" child can hang around
forever. Since the lock file doesn't get cleaned up, this prevents
other users from accessing the target device.

(phk adds: Man, this has been bugging me for YEARS!)

PR: 12528
Submitted by: Craig Leres leres@ee.lbl.gov
MFC after: 1 week

23 years agoAdd a -A option to ping which beeps when packets are lost.
phk [Thu, 24 May 2001 09:17:52 +0000 (09:17 +0000)]
Add a -A option to ping which beeps when packets are lost.

PR: 11818
Submitted by: Marc Evans marc@destek.net

23 years agoBackout rev. 1.22 as the change that caused problems was also backed out.
ru [Thu, 24 May 2001 08:47:57 +0000 (08:47 +0000)]
Backout rev. 1.22 as the change that caused problems was also backed out.

23 years agoMake the rcsid and FreeBSD IDs more sane in the wcs* and wmem* files.
obrien [Thu, 24 May 2001 08:47:42 +0000 (08:47 +0000)]
Make the rcsid and FreeBSD IDs more sane in the wcs* and wmem* files.
Do the same for the non-wcs*/wmem* files while I'm here.

23 years agoThis patch implements O_DIRECT about 80% of the way. It takes a patchset
dillon [Thu, 24 May 2001 07:22:27 +0000 (07:22 +0000)]
This patch implements O_DIRECT about 80% of the way.  It takes a patchset
Tor created a while ago, removes the raw I/O piece (that has cache coherency
problems), and adds a buffer cache / VM freeing piece.

Essentially this patch causes O_DIRECT I/O to not be left in the cache, but
does not prevent it from going through the cache, hence the 80%.  For
the last 20% we need a method by which the I/O can be issued directly to
buffer supplied by the user process and bypass the buffer cache entirely,
but still maintain cache coherency.

I also have the code working under -stable but the changes made to sys/file.h
may not be MFCable, so an MFC is not on the table yet.

Submitted by: tegge, dillon

23 years agoFix $FreeBSD$ style committer messed up in rev 1.7 for some reason.
obrien [Thu, 24 May 2001 07:22:08 +0000 (07:22 +0000)]
Fix $FreeBSD$ style committer messed up in rev 1.7 for some reason.

23 years agoUse bus_space functions rather than inb/outb.
imp [Thu, 24 May 2001 06:54:48 +0000 (06:54 +0000)]
Use bus_space functions rather than inb/outb.
Add defines for PCIC_INDEX and PCIC_DATA offsets.
Change PCIC_INDEX_0 to PCIC_PORT_0
Add define for PCIC_NPORT.
Document why the vadem probe works.

23 years agoProduce a config-time warning about EXT2FS and GPL_MATH_EMULATE
peter [Thu, 24 May 2001 06:26:18 +0000 (06:26 +0000)]
Produce a config-time warning about EXT2FS and GPL_MATH_EMULATE

23 years agoRemove DEV_SNP -> opt_snp.h
peter [Thu, 24 May 2001 06:24:06 +0000 (06:24 +0000)]
Remove DEV_SNP -> opt_snp.h

Forgotten by:  dd

23 years agoDo what we should have done a long time ago:
imp [Thu, 24 May 2001 04:24:22 +0000 (04:24 +0000)]
Do what we should have done a long time ago:

o If the class is PCIC_BRIDGE, subclass is PCIS_BRIDGE_PCMCIA and
  programming interface is 0, assume that it is a generic PCMCIA PCI
  chip we can program.  I don't think there are any of these that
  we don't know about, but you never know.
o If the class is PCIC_BRIDGE, subclass is PCIS_BRIDGE_CARDBUS and
  programming interface is 0, assume that it is a YENTA cardbus bridge
  that we know how to cope with.  There are likely some cardbus bridges
  that haven't it made it in here yet.

23 years agoMove getb1 and putb1 from pcic_isa.c to pcic.c. Rename them to
imp [Thu, 24 May 2001 04:03:28 +0000 (04:03 +0000)]
Move getb1 and putb1 from pcic_isa.c to pcic.c.  Rename them to
pcic_{get,put}b_io.  There are some pci bridges (the CL-PD6729 and
maybe others) that do not have memory mapped registers, so we'll need
these in both places.  Declare them in pcicvar.h.

23 years agoSpelling police: insure -> ensure
dd [Thu, 24 May 2001 03:52:32 +0000 (03:52 +0000)]
Spelling police: insure -> ensure

PR: 27600
Submitted by: Iain Templeton <iain@ugh.net.au>

23 years agoUse pw(8) instead of grep'ing through /etc/passwd. This fixes
dd [Thu, 24 May 2001 03:50:55 +0000 (03:50 +0000)]
Use pw(8) instead of grep'ing through /etc/passwd.  This fixes
automatic filling of the "Originator" field for NIS users.

PR: 24372

23 years agoAdd snp_olddisc member to struct snoop; the linedisc'ized snp(4) uses this.
dd [Thu, 24 May 2001 00:43:44 +0000 (00:43 +0000)]
Add snp_olddisc member to struct snoop; the linedisc'ized snp(4) uses this.

Pointy hat to: dd

23 years agoOops, forgot the 'u' in the getopt for the previous commit.
dillon [Thu, 24 May 2001 00:14:19 +0000 (00:14 +0000)]
Oops, forgot the 'u' in the getopt for the previous commit.

23 years agoCorrect style bugs with regards to long lines and comments.
dd [Wed, 23 May 2001 23:38:05 +0000 (23:38 +0000)]
Correct style bugs with regards to long lines and comments.

Reviewed by: bde

23 years agoBuild snp(4) as a module.
dd [Wed, 23 May 2001 23:33:08 +0000 (23:33 +0000)]
Build snp(4) as a module.

23 years ago*sigh* We can't remove VINUMDEBUG entirely, since we include kernel
grog [Wed, 23 May 2001 23:27:08 +0000 (23:27 +0000)]
*sigh* We can't remove VINUMDEBUG entirely, since we include kernel
header files and sources which depend on it.  For userland, define
VINUMDEBUG here.  Also remove a now superfluous #ifdef.

23 years agoRemove cruft.
grog [Wed, 23 May 2001 23:25:50 +0000 (23:25 +0000)]
Remove cruft.

23 years agoChange #if VINUMDEBUG to #ifdef VINUMDEBUG. This is a flag, not a variable.
grog [Wed, 23 May 2001 23:24:05 +0000 (23:24 +0000)]
Change #if VINUMDEBUG to #ifdef VINUMDEBUG.  This is a flag, not a variable.

23 years agodefine _KERNEL before including <sys/conf.h>
phk [Wed, 23 May 2001 23:03:00 +0000 (23:03 +0000)]
define _KERNEL before including <sys/conf.h>

23 years agoWe don't need to include <sys/conf.h>
phk [Wed, 23 May 2001 23:01:37 +0000 (23:01 +0000)]
We don't need to include <sys/conf.h>

23 years agoDon't acquire Giant just to call trap_fatal(), we are about to panic
jhb [Wed, 23 May 2001 22:58:09 +0000 (22:58 +0000)]
Don't acquire Giant just to call trap_fatal(), we are about to panic
anyway so we'd rather see the printf's then block if the system is
hosed.

23 years agoDon't release Giant around vm_oject_page_clean() in fsync() as the pager
jhb [Wed, 23 May 2001 22:55:13 +0000 (22:55 +0000)]
Don't release Giant around vm_oject_page_clean() in fsync() as the pager
putpages called will need Giant.

23 years agoA feature to allow one to telnet to a unix domain socket. (MFC from
dillon [Wed, 23 May 2001 22:54:07 +0000 (22:54 +0000)]
A feature to allow one to telnet to a unix domain socket. (MFC from
non-crypto version)

Also update the crypto telnet's man page to reflect other options
ported from the non-crypto version.

Obtained from:   Lyndon Nerenberg <lyndon@orthanc.ab.ca>

23 years ago- Assert Giant is held in the vnode pager methods.
jhb [Wed, 23 May 2001 22:51:23 +0000 (22:51 +0000)]
- Assert Giant is held in the vnode pager methods.
- Lock the VM while walking down a vm_object's backing_object list in
  vnode_pager_lock().

23 years ago- Add in several asserts of vm_mtx.
jhb [Wed, 23 May 2001 22:48:28 +0000 (22:48 +0000)]
- Add in several asserts of vm_mtx.
- Assert Giant in vm_pageout_scan() for the vnode hacking that it does.
- Don't hold vm_mtx around vget() or vput().
- Lock Giant when calling vm_pageout_scan() from the pagedaemon.  Also,
  lock curproc while setting the P_BUFEXHAUST flag.
- For now we still hold Giant for all of the vm_daemon.  When process
  limits are locked we will be only need Giant for swapout_procs().

23 years agoA feature to allow one to telnet to a unix domain socket.
dillon [Wed, 23 May 2001 22:47:16 +0000 (22:47 +0000)]
A feature to allow one to telnet to a unix domain socket.

Obtained from:   Lyndon Nerenberg <lyndon@orthanc.ab.ca>

23 years ago- Assert that the vm lock is held for all of _vm_object_allocate().
jhb [Wed, 23 May 2001 22:42:10 +0000 (22:42 +0000)]
- Assert that the vm lock is held for all of _vm_object_allocate().
- Restore the previous order of setting up a new vm_object.  The previous
  had a small bug where we zero'd out the flags after we set the
  OBJ_ONEMAPPING flag.
- Add several asserts of vm_mtx.
- Assert Giant is held rather than locking and unlocking it in a few
  places.
- Add in some #ifdef objlocks code to lock individual vm objects when
  vm objects each have their own lock someday.
- Don't bother acquiring the allproc lock for a ddb command.  If DDB
  blocked on the lock, that would be worse than having an inconsistent
  allproc list.

23 years ago- Add lots of vm_mtx assertions.
jhb [Wed, 23 May 2001 22:38:00 +0000 (22:38 +0000)]
- Add lots of vm_mtx assertions.
- Add a few KTR tracepoints to track the addition and removal of
  vm_map_entry's and the creation adn free'ing of vmspace's.
- Adjust a few portions of code so that we update the process' vmspace
  pointer to its new vmspace before freeing the old vmspace.

23 years ago- Lock the VM around the pmap_swapin_proc() call in faultin().
jhb [Wed, 23 May 2001 22:35:45 +0000 (22:35 +0000)]
- Lock the VM around the pmap_swapin_proc() call in faultin().
- Don't lock Giant in the scheduler() function except for when calling
  faultin().
- In swapout_procs(), lock the VM before the proccess to avoid a lock order
  violation.
- In swapout_procs(), release the allproc lock before calling swapout().
  We restart the process scan after swapping out a process.
- In swapout_procs(), un #if 0 the code to bump the vmspace reference count
  and lock the process' vm structures.  This bug was introduced by me and
  could result in the vmspace being free'd out from under a running
  process.
- Fix an old bug where the vmspace reference count was not free'd if we
  failed the swap_idle_threshold2 test.

23 years ago- Fix the sw_alloc_interlock to actually lock itself when the lock is
jhb [Wed, 23 May 2001 22:31:15 +0000 (22:31 +0000)]
- Fix the sw_alloc_interlock to actually lock itself when the lock is
  acquired.
- Assert Giant is held in the strategy, getpages, and putpages methods and
  the getchainbuf, flushchainbuf, and waitchainbuf functions.
- Always call flushchainbuf() w/o the VM lock.

23 years agoAssert Giant is held for the device pager alloc and getpages methods since
jhb [Wed, 23 May 2001 22:27:52 +0000 (22:27 +0000)]
Assert Giant is held for the device pager alloc and getpages methods since
we call the mmap method of the cdevsw of the device we are mmap'ing.

23 years agoAssert Giant is held by the caller rather than getting it and releasing
jhb [Wed, 23 May 2001 22:26:05 +0000 (22:26 +0000)]
Assert Giant is held by the caller rather than getting it and releasing
it in getpages/putpages.

23 years ago- Always call bfreekva() w/o vm_mtx held.
jhb [Wed, 23 May 2001 22:24:49 +0000 (22:24 +0000)]
- Always call bfreekva() w/o vm_mtx held.
- Always call vfs_setdirty() with vm_mtx held.
- Fix an old comment: vm_hold_unload_pages is called vm_hold_free_pages()
  nowadays.
- Always call vm_hold_free_pages() w/o vm_mtx held.

23 years agoDon't acquire/release Giant around some of the places that need it in
jhb [Wed, 23 May 2001 22:20:29 +0000 (22:20 +0000)]
Don't acquire/release Giant around some of the places that need it in
spec_getpages().  Instead, assert that Giant is held by the caller.

23 years ago- Obtain Giant in mmap() syscall while messing with file descriptors and
jhb [Wed, 23 May 2001 22:17:43 +0000 (22:17 +0000)]
- Obtain Giant in mmap() syscall while messing with file descriptors and
  vnodes.
- Fix an old bug that would leak a reference to a fd if the vnode being
  mmap'd wasn't of type VREG or VCHR.
- Lock Giant in vm_mmap() around calls into the VM that can call into
  pager routines that need Giant or into other VM routines that need
  Giant.
- Replace code that used a goto to jump around the else branch of a test
  to use an else branch instead.

23 years agopmap_mapdev needs the vm_mtx, aquire it if not already locked
alfred [Wed, 23 May 2001 22:17:28 +0000 (22:17 +0000)]
pmap_mapdev needs the vm_mtx, aquire it if not already locked

23 years agolock vm while playing with pmap
alfred [Wed, 23 May 2001 22:13:58 +0000 (22:13 +0000)]
lock vm while playing with pmap

23 years agoAcquire Giant around vm_map_remove() inside of the obreak() syscall for
jhb [Wed, 23 May 2001 22:13:10 +0000 (22:13 +0000)]
Acquire Giant around vm_map_remove() inside of the obreak() syscall for
vm_object_terminate().

23 years agoRegenerate
wpaul [Wed, 23 May 2001 22:11:25 +0000 (22:11 +0000)]
Regenerate

23 years agoTweak the xmphy driver a little bit based on something I learned about
wpaul [Wed, 23 May 2001 22:10:55 +0000 (22:10 +0000)]
Tweak the xmphy driver a little bit based on something I learned about
the built-in 1000baseX interface in the Level 1 LXT1001 chip. The Level 1
PHY comes up with the isolate bit in the control register set by default,
but it also has the autonegotiate bit set. When you tell the xmphy driver
to select IFM_AUTO mode, it sees that the autoneg bit is already on, and
thus doesn't bother updating the control register. However this means that
the isolate bit is never turned off (unless you manually select 1000baseSX
full or half duplex mode, which does result in the control register being
modified and the ISO bit being turned off).

This subtle and unusual behavioral difference stopped me from being able
to receive packets on the SMC9462TX card for several days, since isolating
the PHY disconnects it from the MAC's data interface. The fix is to omit
the 'is the autoneg big set?' test, since it doesn't really provide much
of an optimization anyway.

This commit also updates the xmphy driver to support the Jato/Level 1
internal PHY. (I'm not sure how Jato Technologies is related to Level 1:
all I know is the OUI from the PHY ID registers maps to Jato in the OUI
database.) This will be used once I add the if_lge driver to support
the LXT10010 chip.

23 years agoTake a more conservative approach and still lock Giant around VM faults
jhb [Wed, 23 May 2001 22:09:18 +0000 (22:09 +0000)]
Take a more conservative approach and still lock Giant around VM faults
for now.

23 years ago- Lock the VM when initializing the vmspace for proc0.
jhb [Wed, 23 May 2001 22:06:47 +0000 (22:06 +0000)]
- Lock the VM when initializing the vmspace for proc0.
- Don't bother releasing Giant while doing a lookup on the vm_map of
  initproc while starting up init.  We have to grab it again right after
  the lookup anyways.

23 years agoLock the VM while twiddling the vmspace.
jhb [Wed, 23 May 2001 22:05:08 +0000 (22:05 +0000)]
Lock the VM while twiddling the vmspace.

23 years agoWhen a game ends that makes the top 10, the function insert_score in
phk [Wed, 23 May 2001 21:01:01 +0000 (21:01 +0000)]
When a game ends that makes the top 10, the function insert_score in
score.c is called to make the new score file.  But the case for KFIRE
(killed by fire) incorrectly uses strcpy instead of strcat (all the
other cases use strcat).  This puts the string in the wrong place and
corrupts the score file.

PR: 13278
Submitted by: Mark W. Krentel krentel@dreamscape.com

23 years agoIncrement mbstat.m_mpfail, not mbstat.m_mcfail, when m_pullup() fails.
bmilekic [Wed, 23 May 2001 20:44:54 +0000 (20:44 +0000)]
Increment mbstat.m_mpfail, not mbstat.m_mcfail, when m_pullup() fails.
This slipped in accidently a few commits back.

23 years agoFix memory leak.
ume [Wed, 23 May 2001 20:41:17 +0000 (20:41 +0000)]
Fix memory leak.

Submitted by: itojun

23 years agoPolish getnano{up}*time() prototypes.
phk [Wed, 23 May 2001 20:33:54 +0000 (20:33 +0000)]
Polish getnano{up}*time() prototypes.

PR: 15908
Submitted by: Kelly Yancey kbyanc@posi.net

23 years agofdformat prompts for user confirmation. it exits 0 if the user
phk [Wed, 23 May 2001 20:30:45 +0000 (20:30 +0000)]
fdformat prompts for user confirmation.  it exits 0 if the user
doesn't confirm, which needlessly complicates scripts.

PR: 15339
Submitted by: richard winkel rich@math.missouri.edu
MFC after: 1 week

23 years agoSet the phys_pager_alloc_lock to 1 when it is acquired so that it is
jhb [Wed, 23 May 2001 19:52:23 +0000 (19:52 +0000)]
Set the phys_pager_alloc_lock to 1 when it is acquired so that it is
actually locked.

23 years agoDon't release the vm lock just to turn around and grab it again.
jhb [Wed, 23 May 2001 19:51:12 +0000 (19:51 +0000)]
Don't release the vm lock just to turn around and grab it again.

23 years agoMake sure that all resource allocation is handled in the pcib device, not
dfr [Wed, 23 May 2001 19:44:17 +0000 (19:44 +0000)]
Make sure that all resource allocation is handled in the pcib device, not
the chipset. This is already how the multi-hose systems handle resource
allocation and it fixes a bug where dense and bwx memory allocations were
not handled properly.

Reviewed by: gallatin

23 years agoAdd in assertions to ensure that we always call msleep or mawait with
jhb [Wed, 23 May 2001 19:38:26 +0000 (19:38 +0000)]
Add in assertions to ensure that we always call msleep or mawait with
either a timeout or a held mutex to detect unprotected infinite sleeps
that can easily lead to deadlock.

Submitted by: alfred

23 years agoVlan tags are 12 bits, mask off anything above since the chip doesn't
phk [Wed, 23 May 2001 19:25:14 +0000 (19:25 +0000)]
Vlan tags are 12 bits, mask off anything above since the chip doesn't
seem to do so for us.

PR: 27567
Submitted by: Koji HINO hino@ccm.cl.nec.co.jp
MFC after: 1 week

23 years agosyslogd gets kernel log messages only once every 30 seconds or
phk [Wed, 23 May 2001 19:02:50 +0000 (19:02 +0000)]
syslogd gets kernel log messages only once every 30 seconds or
at the top of the minute, whichever comes first.  It seems
logtimeout() is only called once after the kernel log is opened
and then never again after that.  So I guess syslogd only gets
kernel log messages by virtue of syncer(4)'s flushes ...?

PR: 27361
Submitted by: pkern@utcc.utoronto.ca
MFC after: 1 week

23 years agoAdd some error codes that were added in RFC2616.
des [Wed, 23 May 2001 18:52:02 +0000 (18:52 +0000)]
Add some error codes that were added in RFC2616.

23 years agoNote that sys.mk is read before any other file (even Makefile)
obrien [Wed, 23 May 2001 18:21:36 +0000 (18:21 +0000)]
Note that sys.mk is read before any other file (even Makefile)

23 years agoUse '+' not '|' to add PCATCH to tsleep.
phk [Wed, 23 May 2001 17:54:58 +0000 (17:54 +0000)]
Use '+' not '|' to add PCATCH to tsleep.

23 years agoUse the correct enums in struct sysinit.
phk [Wed, 23 May 2001 17:53:24 +0000 (17:53 +0000)]
Use the correct enums in struct sysinit.

23 years agoRemove the empty uscannerioctl() and use noioctl() instead.
phk [Wed, 23 May 2001 17:51:43 +0000 (17:51 +0000)]
Remove the empty uscannerioctl() and use noioctl() instead.

23 years agoChange the way deletes are managed in DEVFS.
phk [Wed, 23 May 2001 17:48:20 +0000 (17:48 +0000)]
Change the way deletes are managed in DEVFS.

This fixes a number of warnings relating to removed cloned devices.

It also makes it possible to recreate deleted devices with
mknod(2).  The major/minor arguments are ignored.

23 years agoDocument new /usr/include/fs include subdirs.
ru [Wed, 23 May 2001 17:22:38 +0000 (17:22 +0000)]
Document new /usr/include/fs include subdirs.

23 years agoChanged 'blocks' to 'usage' because we may not be getting quotas from
pirzyk [Wed, 23 May 2001 17:12:09 +0000 (17:12 +0000)]
Changed 'blocks' to 'usage' because we may not be getting quotas from
a ufs filesystem and it may be reporting in K instead of 512b blocks.
This is true when using a vxfs filesystem (on a solaris box) for instance.

PR: bin/14545
Submitted by: Jim Pirzyk
Reviewed by: jkh
MFC after: 1 week

23 years agoBackout (almost) revision 1.137 changes.
ru [Wed, 23 May 2001 16:49:02 +0000 (16:49 +0000)]
Backout (almost) revision 1.137 changes.

Removal of LSYMSUBDIRS was a regression.

The purpose of LSYMSUBDIRS is to export only those /sys headers in the
SHARED=symlinks case that are also visible in the SHARED=copies case.

Requested by: bde

23 years agocatch up to i386:
gallatin [Wed, 23 May 2001 16:34:07 +0000 (16:34 +0000)]
catch up to i386:
- remove Giant for vm related traps
- don't hold Giant for MP safe syscalls

Reviewed-by: jhb
23 years agoChanged the clnt_perror string in yp_all because it was saying the
pirzyk [Wed, 23 May 2001 15:37:10 +0000 (15:37 +0000)]
Changed the clnt_perror string in yp_all because it was saying the
error came from yp_next

PR: bin/13254
Submitted by: Jim Pirzyk
Reviewed by: jkh
MFC after: 1 week

23 years agoArgh, unbreak SHARED=copied case.
ru [Wed, 23 May 2001 15:28:53 +0000 (15:28 +0000)]
Argh, unbreak SHARED=copied case.

23 years agoRename (after a repo-copy) some mount(8) programs:
ru [Wed, 23 May 2001 14:58:19 +0000 (14:58 +0000)]
Rename (after a repo-copy) some mount(8) programs:

mount_fdesc -> mount_fdescfs
mount_null -> mount_nullfs
mount_portal -> mount_portalfs
mount_umap -> mount_umapfs
mount_union -> mount_unionfs

23 years agoAdd NO_I4B to avoid building/installing isdn4bsd package.
ru [Wed, 23 May 2001 13:32:32 +0000 (13:32 +0000)]
Add NO_I4B to avoid building/installing isdn4bsd package.

Prompted by: Alexandr Listopad <laa@laa.zp.ua>
MFC after: 3 days

23 years agoFix the range of the returned value.
ru [Wed, 23 May 2001 12:37:50 +0000 (12:37 +0000)]
Fix the range of the returned value.

PR: docs/27446
Submitted by: Tadayuki OKADA <tadayuki@mediaone.net>

23 years agoufs_bmaparray() may block on IO, drop vm mutex and aquire Giant when
alfred [Wed, 23 May 2001 10:30:25 +0000 (10:30 +0000)]
ufs_bmaparray() may block on IO, drop vm mutex and aquire Giant when
calling it from the pager routine