]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoSync
Warner Losh [Thu, 21 Aug 2003 18:05:55 +0000 (18:05 +0000)]
Sync

21 years agoSort the vendors into three sections. First section is for those
Warner Losh [Thu, 21 Aug 2003 18:05:35 +0000 (18:05 +0000)]
Sort the vendors into three sections.  First section is for those
vendors that list the vendor ID in the proper byte order.  The second
section is for vendors that get it backwards.  The third is for what
appear to be 'random' ones (although 0xcxxx appears to be coherent
enough that maybe somebody else is assigning those numbers).

21 years agoNew release notes: OFW_NEWPCI, IP_ONESBCAST, ffsinfo(8).
Bruce A. Mah [Thu, 21 Aug 2003 17:52:08 +0000 (17:52 +0000)]
New release notes:  OFW_NEWPCI, IP_ONESBCAST, ffsinfo(8).

Modified release notes:  Tweak rl(4) note and note MFC.

21 years agoAdd a comment saying it might be Paralon
Warner Losh [Thu, 21 Aug 2003 17:51:17 +0000 (17:51 +0000)]
Add a comment saying it might be Paralon

21 years agoPull in trademark entity definitions.
Bruce A. Mah [Thu, 21 Aug 2003 17:51:05 +0000 (17:51 +0000)]
Pull in trademark entity definitions.

21 years agoSync to 1.68
Warner Losh [Thu, 21 Aug 2003 17:50:26 +0000 (17:50 +0000)]
Sync to 1.68

21 years agoCompaq's ID is 0x138. However, it looks like they also released
Warner Losh [Thu, 21 Aug 2003 17:49:50 +0000 (17:49 +0000)]
Compaq's ID is 0x138.  However, it looks like they also released
something with the vendor ID of 0x183.  That could be a typo, or it
could be Paralon Technologies.  Add an entry for Paralon, but don't
use it yet.

21 years agoLINKSYS2 -> BROMAX
Warner Losh [Thu, 21 Aug 2003 17:40:59 +0000 (17:40 +0000)]
LINKSYS2 -> BROMAX

21 years agoUpdate to pccarddevs 1.67
Warner Losh [Thu, 21 Aug 2003 17:40:28 +0000 (17:40 +0000)]
Update to pccarddevs 1.67

21 years agoVendor ID 0x274 is Bromax Communications, not Linksys. Linksys sells
Warner Losh [Thu, 21 Aug 2003 17:39:56 +0000 (17:39 +0000)]
Vendor ID 0x274 is Bromax Communications, not Linksys.  Linksys sells
them as 'Instant Wireless' cards.

21 years agoCorrect the device id's for the VIA Rhine III chips supported by the
Simon L. B. Nielsen [Thu, 21 Aug 2003 17:38:05 +0000 (17:38 +0000)]
Correct the device id's for the VIA Rhine III chips supported by the
vr(4) driver: VT86C105/VT86C105M -> VT6105/VT6105M.

Confirmed by: driver source code and VIA homepage
Reviewed by: bmah
MFC after: 3 days

21 years agoRetrofit of mac_test regression and consistency test module for MAC
Robert Watson [Thu, 21 Aug 2003 17:28:45 +0000 (17:28 +0000)]
Retrofit of mac_test regression and consistency test module for MAC
Framework labels:

- Re-work the label state assertions to use a set of central
  ASSERT_type_LABEL() assertions.

- Test to make sure labels passed to externalize/internalize calls haven't
  been destroyed.

- For access control checks, assert the condition of all labels passed in.

- For life cycle events, assert the condition of all labels passed in.

- Add new entry point implementations for new MAC Framework entry points:
  mac_test_reflect_mbuf_icmp(), mac_test_reflect_mbuf_tcp(),
  mac_test_check_vnode_deleteextattr(), mac_test_check_vnode_listextattr().

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoDSP Solutions, Inc made the XJEM1144 and XJACK ethernet cards. Update
Warner Losh [Thu, 21 Aug 2003 17:27:49 +0000 (17:27 +0000)]
DSP Solutions, Inc made the XJEM1144 and XJACK ethernet cards.  Update
MEGAHERTZ2 to DSPSI.

21 years agoSync to 1.66
Warner Losh [Thu, 21 Aug 2003 17:26:59 +0000 (17:26 +0000)]
Sync to 1.66

21 years agoVendor ID 0x0128 is registered to DSP Solutions, Inc. Megahertz sold
Warner Losh [Thu, 21 Aug 2003 17:26:44 +0000 (17:26 +0000)]
Vendor ID 0x0128 is registered to DSP Solutions, Inc.  Megahertz sold
these under X-Jack and XJEM1144.  Update to reflect this.

21 years agoUpdate to reflect NETGEAR_2 -> INTERSIL and INTERSIL -> INTERSIL2
Warner Losh [Thu, 21 Aug 2003 17:13:33 +0000 (17:13 +0000)]
Update to reflect NETGEAR_2 -> INTERSIL and INTERSIL -> INTERSIL2
changes to pccarddevs.

21 years agoAdd stub entries for new MAC framework entry points:
Robert Watson [Thu, 21 Aug 2003 17:05:36 +0000 (17:05 +0000)]
Add stub entries for new MAC framework entry points:

  mpo_reflect_mbuf_icmp()
  mpo_reflect_mbuf_tcp()
  mpo_check_vnode_deletextattr()
  mpo_check_vnode_listextattr()

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoHook up repo-copied mac_none.4 to mac_stub.4.
Robert Watson [Thu, 21 Aug 2003 16:53:06 +0000 (16:53 +0000)]
Hook up repo-copied mac_none.4 to mac_stub.4.

Update mac_none to reflect change in role; xref mac_stub.

Update mac_stub to reflect name change, arrival in 5.1; xref
mac_none.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoForced commit to observe repo-copy from mac_none.4 to mac_stub.4.
Robert Watson [Thu, 21 Aug 2003 16:49:49 +0000 (16:49 +0000)]
Forced commit to observe repo-copy from mac_none.4 to mac_stub.4.

21 years agoHook up mac_stub to the modules Makefile.
Robert Watson [Thu, 21 Aug 2003 16:48:39 +0000 (16:48 +0000)]
Hook up mac_stub to the modules Makefile.
Hook up mac_stub in files and options.
Reference mac_stub in NOTES.

21 years agoAdd FreeBSD 5.1 and BSD/OS 5.0 to the family tree.
Murray Stokely [Thu, 21 Aug 2003 16:48:38 +0000 (16:48 +0000)]
Add FreeBSD 5.1 and BSD/OS 5.0 to the family tree.

21 years agoSync to 1.65
Warner Losh [Thu, 21 Aug 2003 16:43:27 +0000 (16:43 +0000)]
Sync to 1.65

21 years agoChange INTERSIL to INTERSIL2 for the -1 case. Chances are good that
Warner Losh [Thu, 21 Aug 2003 16:43:10 +0000 (16:43 +0000)]
Change INTERSIL to INTERSIL2 for the -1 case.  Chances are good that
this card has a real ID, but I don't know what it is at this time.

21 years agoRegen for 1.64
Warner Losh [Thu, 21 Aug 2003 16:41:14 +0000 (16:41 +0000)]
Regen for 1.64

21 years agoVendor 0xb is intersil, not Netgear. The MA401RA is just Netgear's
Warner Losh [Thu, 21 Aug 2003 16:40:20 +0000 (16:40 +0000)]
Vendor 0xb is intersil, not Netgear.  The MA401RA is just Netgear's
modle number, and I continue to use it as a place holder until I find
out what Intersil's name for it is.

21 years agoReplace 'OS release date' with 'kernel release date' here, too.
Eivind Eklund [Thu, 21 Aug 2003 16:36:54 +0000 (16:36 +0000)]
Replace 'OS release date' with 'kernel release date' here, too.

Reminded by: ru

21 years agoGenerally rename things to represent the fact that this is now the
Robert Watson [Thu, 21 Aug 2003 16:22:52 +0000 (16:22 +0000)]
Generally rename things to represent the fact that this is now the
mac_stub policy and no longer mac_none (as found in the repocopy).
Add comment to this effect.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agomac_none is now the null policy, not a stub policy, so remove the
Robert Watson [Thu, 21 Aug 2003 16:19:17 +0000 (16:19 +0000)]
mac_none is now the null policy, not a stub policy, so remove the
stubs.  Add a pointer to mac_stub, which is now the stub policy.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoApply Rudolf Cejka's patch for:
Orion Hodson [Thu, 21 Aug 2003 15:50:32 +0000 (15:50 +0000)]
Apply Rudolf Cejka's patch for:

o AD1980 hook.
o ac97_fix_auxout.

and:

o Associate AC97_MIX_AUXOUT with SOUND_MIXER_OGAIN rather than
SOUND_MIXER_MONITOR.

o Add ac97_fix_tone to remove tone controls from mixer if invalid.

21 years agoAdd patch for AD198x.
Orion Hodson [Thu, 21 Aug 2003 15:44:55 +0000 (15:44 +0000)]
Add patch for AD198x.

Submitted by: Oleg Sharoiko, Rudolf Cejka.

21 years agoForced commit to reflect repo-copy of:
Robert Watson [Thu, 21 Aug 2003 15:43:01 +0000 (15:43 +0000)]
Forced commit to reflect repo-copy of:

  modules/mac_none/Makefile -> modules/mac_stub/Makefile
  security/mac_none/mac_none.c -> security/mac_stub/mac_stub.c

mac_none will become a null policy, and mac_stub will be the stub
policy from here on out.

Thanks Joe!

21 years agoCatch up ep(4) with the hardware notes.
Simon L. B. Nielsen [Thu, 21 Aug 2003 15:40:05 +0000 (15:40 +0000)]
Catch up ep(4) with the hardware notes.

Submitted by: Lukas Ertl <l.ertl@univie.ac.at>
PR: docs/55659
Confirmed by: driver source code and pccard.conf
MFC after: 3 days

21 years agoCatch up fe(4) with the hardware notes.
Simon L. B. Nielsen [Thu, 21 Aug 2003 15:18:21 +0000 (15:18 +0000)]
Catch up fe(4) with the hardware notes.

Submitted by: Lukas Ertl <l.ertl@univie.ac.at>
PR: docs/55636
Confirmed by: driver source code and pccard.conf
MFC after: 3 days

21 years agoChange description of kern.osreldate from "Operating system release date" to
Eivind Eklund [Thu, 21 Aug 2003 14:47:08 +0000 (14:47 +0000)]
Change description of kern.osreldate from "Operating system release date" to
"Kernel release date" - userland version is in /usr/include/osreldate.h

21 years agoImplementations of mpo_check_vnode_deleteextattr() and
Robert Watson [Thu, 21 Aug 2003 14:34:54 +0000 (14:34 +0000)]
Implementations of mpo_check_vnode_deleteextattr() and
mpo_check_vnode_listextattr() for Biba, MLS, and BSD Extended.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoAdd the SmartBridges SmartLink to the list of supported devices.
Simon L. B. Nielsen [Thu, 21 Aug 2003 14:26:29 +0000 (14:26 +0000)]
Add the SmartBridges SmartLink to the list of supported devices.

Submitted by: Lukas Ertl <l.ertl@univie.ac.at>
PR: docs/55644
Confirmed by: driver source code
MFC after: 3 days

21 years agoRegen.
John Baldwin [Thu, 21 Aug 2003 14:16:41 +0000 (14:16 +0000)]
Regen.

21 years agoSwap sigaction/sigreturn since they are in the wrong order.
John Baldwin [Thu, 21 Aug 2003 14:16:00 +0000 (14:16 +0000)]
Swap sigaction/sigreturn since they are in the wrong order.

Noticed indirectly by: peter

21 years agoAdd mac_check_vnode_deleteextattr() and mac_check_vnode_listextattr():
Robert Watson [Thu, 21 Aug 2003 13:53:01 +0000 (13:53 +0000)]
Add mac_check_vnode_deleteextattr() and mac_check_vnode_listextattr():
explicit access control checks to delete and list extended attributes
on a vnode, rather than implicitly combining with the setextattr and
getextattr checks.  This reflects EA API changes in the kernel made
recently, including the move to explicit VOP's for both of these
operations.

Obtained from: TrustedBSD PRoject
Sponsored by: DARPA, Network Associates Laboratories

21 years agoInclude order bit in sdev->type.
Hidetoshi Shimokawa [Thu, 21 Aug 2003 08:40:31 +0000 (08:40 +0000)]
Include order bit in sdev->type.

21 years agoAdd in a missing device entry that I thought was already in there. This
Doug Ambrisko [Thu, 21 Aug 2003 03:54:20 +0000 (03:54 +0000)]
Add in a missing device entry that I thought was already in there.  This
is from the same PR.  Prompted by the reporter of the PR.

PR: 38372

21 years agoRegen
Peter Wemm [Thu, 21 Aug 2003 03:48:50 +0000 (03:48 +0000)]
Regen

21 years agoThis is too funny for words. Swap syscalls 416 and 417 around. It works
Peter Wemm [Thu, 21 Aug 2003 03:48:05 +0000 (03:48 +0000)]
This is too funny for words.  Swap syscalls 416 and 417 around.  It works
better that way when sigaction() and sigreturn() do the right thing.

21 years agoMinimal update to make it easier to increase the buffer-size lpd uses
Garance A Drosehn [Thu, 21 Aug 2003 03:43:48 +0000 (03:43 +0000)]
Minimal update to make it easier to increase the buffer-size lpd uses
when reading/writing spool files.  I intend to do a more elaborate
version, but I want to get this much in before 4.9-release.  As written,
this results in no change to the object code.

Submitted by: John-Mark Gurney
Reviewed by: /sbin/md5
MFC after: 4 days

21 years agoIn my efforts to make checking the CD bits less error-prone, I
Warner Losh [Wed, 20 Aug 2003 23:43:20 +0000 (23:43 +0000)]
In my efforts to make checking the CD bits less error-prone, I
introduced an error.  In power_enable_socket, we only return ENODEV if
there isn't a card, not if there is a card.  Ooops.

Pointy hat to: bsdimp

21 years agoWrite the short and long version of the month
Mike Makonnen [Wed, 20 Aug 2003 22:35:55 +0000 (22:35 +0000)]
Write the short and long version of the month
names in Amharic instead of English.
Also, remove some extra names I had previously considered
including. They don't make sense since the calendar names
don't match up.

21 years agoAdd Protocol Independent Multicast protocol.
Jeffrey Hsu [Wed, 20 Aug 2003 22:11:58 +0000 (22:11 +0000)]
Add Protocol Independent Multicast protocol.

Submitted by: Pavlin Radoslavov <pavlin@icir.org>

21 years agoAssert that the vm object lock is held in vm_page_alloc().
Alan Cox [Wed, 20 Aug 2003 20:24:29 +0000 (20:24 +0000)]
Assert that the vm object lock is held in vm_page_alloc().

21 years ago - Lock the pte object when performing vm_page_grab().
Alan Cox [Wed, 20 Aug 2003 20:12:05 +0000 (20:12 +0000)]
 - Lock the pte object when performing vm_page_grab().
 - Insure that the page table page is zero filled before adding it
   to the page table.

21 years agoRemove about 40 lines of #ifdef/#endif by using new macros
Robert Watson [Wed, 20 Aug 2003 19:16:49 +0000 (19:16 +0000)]
Remove about 40 lines of #ifdef/#endif by using new macros
MAC_DEBUG_COUNTER_INC() and MAC_DEBUG_COUNTER_DEC() to maintain
debugging counter values rather than #ifdef'ing the atomic
operations to MAC_DEBUG.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agobde made a number of suggested improvements to the code. This commit
Warner Losh [Wed, 20 Aug 2003 19:12:46 +0000 (19:12 +0000)]
bde made a number of suggested improvements to the code.  This commit
represents the pruely stylistic changes and should have no net impact
on the rest of the code.

bde's more substantive changes will follow in a separate commit once
we've come to closure on them.

Submitted by: bde

21 years agoIn sysctl_vm_zone, do not calculate per-cpu cache stats on
Bosko Milekic [Wed, 20 Aug 2003 18:22:06 +0000 (18:22 +0000)]
In sysctl_vm_zone, do not calculate per-cpu cache stats on
UMA_ZFLAG_INTERNAL zones at all.  Apparently, Wilko's alpha
was crashing while entering multi-user because, I think, we
were calculating the garbage cachefree for pcpu caches that
essentially don't exist for at least the 'zones' zone and it so
happened that we were reading from an unmapped location.

Confirmed to fix crash: wilko
Helped debug: wilko, gallatin

21 years agoBefore digging into IGMP locking, do a whitespace and prototype cleanup:
Robert Watson [Wed, 20 Aug 2003 17:32:17 +0000 (17:32 +0000)]
Before digging into IGMP locking, do a whitespace and prototype cleanup:
prefer tabs to 8 spaces, focus on consistent indentation, prefer modern
C function prototypes.  Not all the way to style(9), but substantially
closer.

21 years agoMove from a custom-crafted singly-linked list to the SLIST_* macros
Robert Watson [Wed, 20 Aug 2003 17:09:01 +0000 (17:09 +0000)]
Move from a custom-crafted singly-linked list to the SLIST_* macros
from queue(3).

Improve vertical compactness by using a IGMP_PRINTF() macro rather
than #ifdefing IGMP_DEBUG a large number of debugging printfs.

Reviewed by: mdodd (SLIST changes)

21 years agoFix accidental line deletion.
David E. O'Brien [Wed, 20 Aug 2003 15:42:05 +0000 (15:42 +0000)]
Fix accidental line deletion.

21 years agoAdd the IP_ONESBCAST option, to enable undirected IP broadcasts to be sent on
Bruce M Simpson [Wed, 20 Aug 2003 14:46:40 +0000 (14:46 +0000)]
Add the IP_ONESBCAST option, to enable undirected IP broadcasts to be sent on
specific interfaces. This is required by aodvd, and may in future help us
in getting rid of the requirement for BPF from our import of isc-dhcp.

Suggested by:   fenestro
Obtained from:  BSD/OS
Reviewed by:    mini, sam
Approved by:    jake (mentor)

21 years ago_thr_sig_check_pending is also called by scope system thread when it leaves
David Xu [Wed, 20 Aug 2003 13:43:35 +0000 (13:43 +0000)]
_thr_sig_check_pending is also called by scope system thread when it leaves
critical region, we wrap some syscalls for thread cancellation point, and
when syscalls returns, we call _thr_leave_cancellation_point, at the time
if a signal comes in, it would be buffered, and when the thread leaves
_thr_leave_cancellation_point, buffered signals will be processed, to avoid
messing up normal syscall errno, we should save and restore errno around
signal handling code.

21 years agoFix typo in last commit.
Maxime Henrion [Wed, 20 Aug 2003 11:27:58 +0000 (11:27 +0000)]
Fix typo in last commit.

21 years agoGet rid of a __DECONST by strdup'ing the string in question. When
Hartmut Brandt [Wed, 20 Aug 2003 08:25:36 +0000 (08:25 +0000)]
Get rid of a __DECONST by strdup'ing the string in question. When
called this way the program just prints its help intro, so the
memory leak is not a problem.

Pointed out by: bde

21 years agoFix build by correcting a wrong cast.
Maxime Henrion [Wed, 20 Aug 2003 08:24:49 +0000 (08:24 +0000)]
Fix build by correcting a wrong cast.

21 years agoo Reduce rc(8) startup clutter by turning the informational messages
Mike Makonnen [Wed, 20 Aug 2003 06:50:34 +0000 (06:50 +0000)]
o Reduce rc(8) startup clutter by turning the informational messages
  off by default.
o Apparently the routine displaying the informational messages wasn't
  checking its knob in rc.conf, so fix that as well.

Requested by: obrien

21 years agoUse PC Card instead of other variants in the driver man pages.
Warner Losh [Wed, 20 Aug 2003 06:49:57 +0000 (06:49 +0000)]
Use PC Card instead of other variants in the driver man pages.

21 years agoThe PCMCIA Standard dictates that those funny cards you insert into
Warner Losh [Wed, 20 Aug 2003 06:27:21 +0000 (06:27 +0000)]
The PCMCIA Standard dictates that those funny cards you insert into
laptops are "PC Cards" and uses said term consistantly.  Allow my
foolish hobgoblins to get the better of me and become consistant.

21 years agoAdd a new variable 'skipPCCARD'. This variable will cause sysinstall
Warner Losh [Wed, 20 Aug 2003 06:24:12 +0000 (06:24 +0000)]
Add a new variable 'skipPCCARD'.  This variable will cause sysinstall
to ignore all PC Card devices.

Submitted by: Anders Nordby
PR: bin/37650

MFC After: 2 weeks

21 years agoAdd a general mechanism for creating and applying
Mike Makonnen [Wed, 20 Aug 2003 06:15:18 +0000 (06:15 +0000)]
Add a general mechanism for creating and applying
devfs(8) rules in rc(8). It is most useful for applying
rules to devfs(5) mount points in /dev or inside jails.
The following line of script is sufficient to
mount a relatively useful+secure devfs(5) in a jail:

devfs_mount_jail /some/jail/dev

Some new shell routines available to scripts that source
rc.subr(5):
o devfs_link - Makes it a little easier to create symlinks
o devfs_init_rulesets - Create devfs(8) rulesets from devfs.rules
o devfs_set_ruleset - Set a ruleset to a devfs(5) mount
o devfs_apply_ruleset - Apply a ruleset to a devfs(5) mount
o devfs_domount - Mount devfs(5) and apply some ruleset
o devfs_mount_jail - Mount devfs(5) and apply a ruleset
  appropriate to jails.

Additional rulesets can be specified in /etc/devfs.rules.
If the devfs_system_ruleset variable is defined in rc.conf
and it contains the name of a ruleset defined in /etc/defaults/devfs.rules
or user supplied rulesets in /etc/devfs.rules then that ruleset will
be applied to /dev at startup by the /etc/rc.d/devfs script. It can
also be applied post-startup:

/etc/rc.d/devfs start

This is a more flexible mechanism than the previous method of using
/etc/devfs.conf. However, that method is still available.

Note: since devfs(8) doesn't provide any way for creating symlinks
as part of a ruleset, anyone wishing to create symlinks in a devfs(5)
as part of the bootup sequence will still have to rely on /etc/devfs.conf.

21 years agoUse CBB_CARD_PRESENT instead of checking STATE_CD directly since it is
Warner Losh [Wed, 20 Aug 2003 06:13:01 +0000 (06:13 +0000)]
Use CBB_CARD_PRESENT instead of checking STATE_CD directly since it is
hard to get right.

21 years agoLock the pmap's tsb object when performing vm_page_grab() on it.
Alan Cox [Wed, 20 Aug 2003 06:11:39 +0000 (06:11 +0000)]
Lock the pmap's tsb object when performing vm_page_grab() on it.

21 years agoFix panic from having multiple locks held when updating firmware. AN_LOCK
Doug Ambrisko [Wed, 20 Aug 2003 05:52:48 +0000 (05:52 +0000)]
Fix panic from having multiple locks held when updating firmware.  AN_LOCK
was asserted during a tsleep resulting in a panic in tsleep.

21 years agopc98pcic isn't the name of the media driver. tcic likely will never
Warner Losh [Wed, 20 Aug 2003 05:50:56 +0000 (05:50 +0000)]
pc98pcic isn't the name of the media driver.  tcic likely will never
happen (unless someone sends one to me).  Kill bogus module depend
that I commented out over a year ago.

21 years agoWhen debugging CIS, only print 10 CISTPL_NULLs. Chances are good they
Warner Losh [Wed, 20 Aug 2003 05:44:55 +0000 (05:44 +0000)]
When debugging CIS, only print 10 CISTPL_NULLs.  Chances are good they
are all bogus, and the cards that don't decode things quite right
often have hundreds of them.  This will fix starvation of small dmesg
buffers and allow better debugging to happen.  I thought about adding
an override, but there is such a thing as too many knobs. :-)

21 years agoFix an extreme edge case in leap second handling. We need to call
Warner Losh [Wed, 20 Aug 2003 05:34:27 +0000 (05:34 +0000)]
Fix an extreme edge case in leap second handling.  We need to call
ntp_update_second twice when we have a large step in case that step
goes across a scheduled leap second.  The only way this could happen
would be if we didn't call tc_windup over the end of day on the day of
a leap second, which would only happen if timeouts were delayed for
seconds.  While it is an edge case, it is an important one to get
right for my employer.

Sponsored by: Timing Solutions Corporation

21 years agoUndo the mistake made in revision 1.77 of trap.c and which was the
Marcel Moolenaar [Wed, 20 Aug 2003 05:30:35 +0000 (05:30 +0000)]
Undo the mistake made in revision 1.77 of trap.c and which was the
ultimate trigger for the follow-up fixes in revisions 1.78, 1.80,
1.81 and 1.82 of trap.c. I was simply too pre-occupied with the
gateway page and how it blurs kernel space with user space and
vice versa that I couldn't see that it was all a load of bollocks.

It's not the IP address that matters, it's the privilege level that
counts. We never run in user space with lifted permissions and we
sure can not run in kernel space without it. Sure, the gateway page
is the exception, but not if you look at the privilege level. It's
user space if you run with user permissions and kernel space otherwise.

So, we're back to looking at the privilege level like it should be.
There's no other way.

Pointy hat: marcel

21 years ago - Lock the pte object when performing vm_page_grab().
Alan Cox [Wed, 20 Aug 2003 05:09:55 +0000 (05:09 +0000)]
 - Lock the pte object when performing vm_page_grab().
 - Insure that the page table page is zero filled before adding it
   to the page table.

21 years agoOn Dell boxes such as the PE2650 we need to disable 3 led mode for the
Doug Ambrisko [Wed, 20 Aug 2003 04:06:00 +0000 (04:06 +0000)]
On Dell boxes such as the PE2650 we need to disable 3 led mode for the
link LED to work.

Reviewed by: wpaul
Obtained from: Linux driver

21 years ago- Add support for Cisco latest firmware RID sizes that supports 25 SSIDs!
Doug Ambrisko [Wed, 20 Aug 2003 03:46:05 +0000 (03:46 +0000)]
- Add support for Cisco latest firmware RID sizes that supports 25 SSIDs!
- Fix up TX speed changes.
- Make mpi-350 cards sort-of work with new firmware.  It RXs okay but TXs
  only work for about 14 packets then fails to get an interrupt.  The
  TX watchdog fires.  It has been reported that my hack for now doesn't
  break cards with the older firmware.  It appears my card has lost
  the ability to RX or TX at all but other peoples cards work.  I assume
  it got damaged in tansport.

MFC: 1 week.

21 years agoFill reserved fields of transmitting packets header with zero.
Hidetoshi Shimokawa [Wed, 20 Aug 2003 03:11:37 +0000 (03:11 +0000)]
Fill reserved fields of transmitting packets header with zero.
This fixes the if_fwe problem with 1394b chip.

PR and Tested by: nork

21 years agoAdd back a loop for up to PTHREAD_DESTRUCTOR_ITERATIONS to
Daniel Eischen [Wed, 20 Aug 2003 02:34:14 +0000 (02:34 +0000)]
Add back a loop for up to PTHREAD_DESTRUCTOR_ITERATIONS to
destroy thread-specific data.  Display a warning when thread
specific data remains after PTHREAD_DESTRUCTOR_ITERATIONS.

Reviewed by: davidxu

21 years agoDocument the OFW_NEWPCI commit to sparc64 generic.
David E. O'Brien [Wed, 20 Aug 2003 01:52:30 +0000 (01:52 +0000)]
Document the OFW_NEWPCI commit to sparc64 generic.

21 years agoRemove the vestiges of the old pre-"X_AS_PKG" way we used to handled the
David E. O'Brien [Tue, 19 Aug 2003 23:23:27 +0000 (23:23 +0000)]
Remove the vestiges of the old pre-"X_AS_PKG" way we used to handled the
installing XFree86 (version 3.3.6 and before).

Reviewed by: jhb

21 years agoAdd a kluge suggested by Marcel to paper over the difference between
Garrett Wollman [Tue, 19 Aug 2003 23:01:46 +0000 (23:01 +0000)]
Add a kluge suggested by Marcel to paper over the difference between
gethostname()'s old and new signatures without requiring a library
bump.  Note that programs which called gethostname() with a negative
argument were already broken, since the same type conversion was done
by the old implementation.  Add a note in the Makefile so that whoever
next bumps the libc revision will delete the kluge at the same time
(as it will no longer be necessary).  This is only operative on 64-bit
platforms.

Submitted by: marcel

21 years agoMFp4 changes to fix locking issues and correct reference
Sam Leffler [Tue, 19 Aug 2003 22:17:04 +0000 (22:17 +0000)]
MFp4 changes to fix locking issues and correct reference
count handling of station entries in hostap mode:

Input path:

o driver is now expected to find the node associated with the
  sender of a received frame; use ic_bss if none is located
o driver passes the (referenced) node into ieee80211_input for
  use within the wlan module and is responsible for cleaning up
  on return
o the antenna state is no longer passed up with each frame; this
  is now considered driver-private state and drivers are responsible
  for keeping it in the driver-private part of a node

Output path:

Revamp output path for management frames to eliminate redundant
locking that causes problems and to correct reference counting
bogosity that occurs when stations are timed out due to inactivity
(in AP mode).  On output the refcnt'd node is stashed in the pkthdr's
recvif field (yech) and retrieved by the driver.  This eliminates
an unref/ref scenario and related node table unlock/lock due to the
driver looking up the node.  This is particularly important when
stations are timed out as this causes a lock order reversal that
can result in a deadlock.  As a byproduct we also reduce the overhead
for sending management frames (minimal).  Additional fallout from
this is a change to ieee80211_encap to return a refcn't node for
tieing to the outbound frame.  Node refcnts are not reclaimed until
after a frame is completely processed (e.g. in the tx interrupt
handler).  This is especially important for timed out stations as
this deref will be the final one causing the node entry to be
reclaimed.

Additional semi-related changes:
o replace m_copym use with m_copypacket (optimization)
o add assert to verify ic_bss is never free'd during normal operation
o add comments explaining calling conventions by drivers for frames
  going in each direction
o remove extraneous code that "cannot be executed" (e.g. because
  pointers may never be null)

21 years agoEnable OFW_NEWPCI until jmg's 2003/06/21 18:26:08 PDT bus commit is fixed
David E. O'Brien [Tue, 19 Aug 2003 21:57:29 +0000 (21:57 +0000)]
Enable OFW_NEWPCI until jmg's 2003/06/21 18:26:08 PDT bus commit is fixed
that caused a 3-4 times slow down in performance.
(the primary Sparc64 developers are all using OFW_NEWPCI already, so it is
the best code path for users)

21 years agoo pass control frames up the stack when in monitor mode (the 802.11 layer will
Sam Leffler [Tue, 19 Aug 2003 21:35:08 +0000 (21:35 +0000)]
o pass control frames up the stack when in monitor mode (the 802.11 layer will
  quietly discard them; this just permits them to be collected with bpf)
o add a counter for the number of rate control frames discarded when not in
  monitor mode
o move the rx "too short" statistic in the stat structure so non-error rx stats
  are together (NB: ABI change to apps that collect stats via driver ioctl)

21 years agoMake aac(4) compile cleanly on 64-bit machines. The code was already 64-bit
Scott Long [Tue, 19 Aug 2003 21:31:45 +0000 (21:31 +0000)]
Make aac(4) compile cleanly on 64-bit machines.  The code was already 64-bit
safe, but some (unneeded and/or harmless) downcasts were generating warnings.
The driver still is not endian-clean.

21 years agoo correct beacon frame length calculation and add an assert to catch any future
Sam Leffler [Tue, 19 Aug 2003 21:28:45 +0000 (21:28 +0000)]
o correct beacon frame length calculation and add an assert to catch any future
  mistakes (this mistake was not an issue because the length is only used to
  decide whether or not to allocate a cluster)
o while here, move a beacon length comment to the "right place"

21 years agomaintain a table for mapping hardware rate codes to 802.11 rates for
Sam Leffler [Tue, 19 Aug 2003 21:24:16 +0000 (21:24 +0000)]
maintain a table for mapping hardware rate codes to 802.11 rates for
calculating the rate for each rx'd frame

21 years agomark the scan and calibrate callouts MPSAFE
Sam Leffler [Tue, 19 Aug 2003 21:05:47 +0000 (21:05 +0000)]
mark the scan and calibrate callouts MPSAFE

21 years agoremove unneeded include files
Sam Leffler [Tue, 19 Aug 2003 21:03:30 +0000 (21:03 +0000)]
remove unneeded include files

21 years agoUpdate gethostname() prototype to match source and standard.
Garrett Wollman [Tue, 19 Aug 2003 20:39:49 +0000 (20:39 +0000)]
Update gethostname() prototype to match source and standard.

21 years agoChange gethostname() to set errno to ENAMETOOLONG instead of ENOMEM
Garrett Wollman [Tue, 19 Aug 2003 20:38:44 +0000 (20:38 +0000)]
Change gethostname() to set errno to ENAMETOOLONG instead of ENOMEM
when the buffer is not long enough to hold the current host name.
POSIX does not standardize error returns for gethostname(), so it
doesn't matter which one we use, but ENAMETOOLONG is at least a little
more intuitive, and mi suggests the existence of prior art.  I've been
running with this change for a while on my home machine with no
effect.  At the same time, I've updated the prototype for
gethostname() to use the correct standard type (size_t) for the
namelen argument.

All of the in-tree callers fall into one of the following categories:
1) Call perror() or equivalent when gethostname() fails.
2) Ignore gethostname()'s return value entirely, potentially resulting
in data corruption if the buffer is too small.
3) Fall back to a (possibly sensible) default value if gethostname()
fails.

Many of the callers I examined shows signs of confusion about the
correct sizing of the host name buffer.  gethostname(3) now has more
information about this, as well as updated standards information.

PR: 48114
Submitted by: mi (in part)

21 years agoEliminate a possible race condition for multithreaded applications in
Alan Cox [Tue, 19 Aug 2003 18:20:34 +0000 (18:20 +0000)]
Eliminate a possible race condition for multithreaded applications in
_pmap_allocpte(): Guarantee that the page table page is zero filled before
adding it to the directory.  Otherwise, a 2nd, 3rd, etc. thread could
access a nearby virtual address and use garbage for the address
translation.

Discussed with: peter, tegge

21 years agoExpand the fdisk size display toggling to include GB.
David E. O'Brien [Tue, 19 Aug 2003 17:51:49 +0000 (17:51 +0000)]
Expand the fdisk size display toggling to include GB.

21 years agoChange instances of callout_init that specify MPSAFE behaviour to
Sam Leffler [Tue, 19 Aug 2003 17:51:11 +0000 (17:51 +0000)]
Change instances of callout_init that specify MPSAFE behaviour to
use CALLOUT_MPSAFE instead of "1" for the second parameter.  This
does not change the behaviour; it just makes the intent more clear.

21 years agoProperly quote the lib path.
David E. O'Brien [Tue, 19 Aug 2003 17:30:46 +0000 (17:30 +0000)]
Properly quote the lib path.

21 years agoadd R_Zalloc definition that returns pre-zero'd memory
Sam Leffler [Tue, 19 Aug 2003 17:23:07 +0000 (17:23 +0000)]
add R_Zalloc definition that returns pre-zero'd memory

21 years ago* Bug fix in bw_meter_process(): the periodically processed bins
Jeffrey Hsu [Tue, 19 Aug 2003 17:22:51 +0000 (17:22 +0000)]
* Bug fix in bw_meter_process(): the periodically processed bins
  of bw_meter entries were processed up to one second ahead.
  After an unappropriate rescheduling of some of the bw_meter
  entries, the upcalls weren't delivered.

* pim_register_prepare() uses the appropriate sw_csum flag to
  call ip_fragment() so the IP checksum is computed properly.

* Modify pim_register_prepare() to take care of IP packets that
  don't need fragmentation.

* Add-back in_delayed_cksum() to encap_send(), because it seems it
  should be there.

Submitted by: Pavlin Radoslavov <pavlin@icir.org>

21 years agoremove #define no longer used
Sam Leffler [Tue, 19 Aug 2003 17:12:21 +0000 (17:12 +0000)]
remove #define no longer used

21 years agoadd missing unlock when in_pcballoc returns an error
Sam Leffler [Tue, 19 Aug 2003 17:11:46 +0000 (17:11 +0000)]
add missing unlock when in_pcballoc returns an error

21 years agouse ETHER_IS_MULTICAST instead of explicit check
Sam Leffler [Tue, 19 Aug 2003 17:04:13 +0000 (17:04 +0000)]
use ETHER_IS_MULTICAST instead of explicit check

21 years ago"short cable" fix for DP8315
Sam Leffler [Tue, 19 Aug 2003 16:57:00 +0000 (16:57 +0000)]
"short cable" fix for DP8315

Derived from: netbsd and tech note at soekris web site
MFC after: 1 week