]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoAdd an example how to use keyfiles for encrypted providers which should be
pjd [Sat, 11 Feb 2006 13:18:47 +0000 (13:18 +0000)]
Add an example how to use keyfiles for encrypted providers which should be
attached before the root file system is mounted.

MFC after: 3 days

18 years agoTeach geli how to load keyfiles before root file system is mounted.
pjd [Sat, 11 Feb 2006 13:08:24 +0000 (13:08 +0000)]
Teach geli how to load keyfiles before root file system is mounted.
An example entries for loader.conf to make it possible:

geli_da0_keyfile0_load="YES"
geli_da0_keyfile0_type="da0:geli_keyfile0"
geli_da0_keyfile0_name="/boot/keys/da0.key0"
geli_da0_keyfile1_load="YES"
geli_da0_keyfile1_type="da0:geli_keyfile1"
geli_da0_keyfile1_name="/boot/keys/da0.key1"
geli_da0_keyfile2_load="YES"
geli_da0_keyfile2_type="da0:geli_keyfile2"
geli_da0_keyfile2_name="/boot/keys/da0.key2"

geli_da1s3a_keyfile0_load="YES"
geli_da1s3a_keyfile0_type="da1s3a:geli_keyfile0"
geli_da1s3a_keyfile0_name="/boot/keys/da1s3a.key"

Thanks for jhb and kan who showed me the right direction.

MFC after: 3 days

18 years ago- Allow to use -b without passphrase or with keyfiles as it will be
pjd [Sat, 11 Feb 2006 13:04:10 +0000 (13:04 +0000)]
- Allow to use -b without passphrase or with keyfiles as it will be
  supported for a moment.
- Don't allow to use -i when no passphrase is given. Now if iterations is
  equal to -1 (not set), we know that we should not ask for the passphrase
  on boot.
  It still doesn't handle situation when one key is protected with
  passphrase and the other is not. There is no quick fix for this.
  The complete solution will be to make number of iterations a per-key
  value. Because this need metadata format change and is only needed for
  devices attached on boot, I'll leave it as it is for now.

MFC after: 3 days

18 years agoCheck rootvnode variable to see if we still want to ask for passphrase on
pjd [Sat, 11 Feb 2006 12:45:01 +0000 (12:45 +0000)]
Check rootvnode variable to see if we still want to ask for passphrase on
boot. Other methods just don't work properly.

MFC after: 3 days

18 years agoCPU time accounting speedup (step 2)
phk [Sat, 11 Feb 2006 09:33:07 +0000 (09:33 +0000)]
CPU time accounting speedup (step 2)

Keep accounting time (in per-cpu) cputicks and the statistics counts
in the thread and summarize into struct proc when at context switch.

Don't reach across CPUs in calcru().

Add code to calibrate the top speed of cpu_tickrate() for variable
cpu_tick hardware (like TSC on power managed machines).

Don't enforce monotonicity (at least for now) in calcru.  While the
calibrated cpu_tickrate ramps up it may not be true.

Use 27MHz counter on i386/Geode.

Use TSC on amd64 & i386 if present.

Use tick counter on sparc64

18 years agoUse the correct address for the ohci device.
imp [Sat, 11 Feb 2006 03:58:07 +0000 (03:58 +0000)]
Use the correct address for the ohci device.

18 years agoUse ETHER_ADDR_LEN rather than 6.
imp [Sat, 11 Feb 2006 03:50:03 +0000 (03:50 +0000)]
Use ETHER_ADDR_LEN rather than 6.

18 years agoImprove the memory resoruce allocation and usage during the probe.
imp [Sat, 11 Feb 2006 03:41:20 +0000 (03:41 +0000)]
Improve the memory resoruce allocation and usage during the probe.
I've had these improvements in my tree for almost a year now...

18 years agoUse the release version as a floor, not an exact match. Add comment to that
imp [Sat, 11 Feb 2006 03:30:17 +0000 (03:30 +0000)]
Use the release version as a floor, not an exact match.  Add comment to that
effect and draw attention to the fact that the list has stronger ordering
requirements than before.

18 years agoGive ohci_intr a proper return value. In FreeBSD, this is void. Remove
imp [Sat, 11 Feb 2006 03:29:02 +0000 (03:29 +0000)]
Give ohci_intr a proper return value.  In FreeBSD, this is void.  Remove
now-redundant cast on establishing the interrupt.

18 years agoUpdate the keyspan product list from NetBSD.
imp [Sat, 11 Feb 2006 03:27:57 +0000 (03:27 +0000)]
Update the keyspan product list from NetBSD.

18 years agoDo initial cut of SAS HBA support. These controllers (106X) seem to support
mjacob [Sat, 11 Feb 2006 01:35:29 +0000 (01:35 +0000)]
Do initial cut of SAS HBA support. These controllers (106X) seem to support
automatically both SATA and SAS drives.  The async SAS event handling we catch
but ignore at present (so automagic attach/detach isn't hooked up yet).

Do 64 bit PCI support- we can now work on systems with > 4GB of memory.

Do large transfer support- we now can support up to reported chain depth, or
the length of our request area. We simply allocate additional request elements
when we would run out of room for chain lists.

Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms.
There were no RAID cards available for me to regression test.

The error recovery for this driver still is pretty bad.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r155518,
rwatson [Sat, 11 Feb 2006 00:39:23 +0000 (00:39 +0000)]
This commit was generated by cvs2svn to compensate for changes in r155518,
which included commits to RCS files with non-trunk default branches.

18 years agoCVS import OpenBSM 1.0 alpha 4:
rwatson [Sat, 11 Feb 2006 00:39:23 +0000 (00:39 +0000)]
CVS import OpenBSM 1.0 alpha 4:

- Remove "audit" user example from audit_user, as it's not present on most
  systems.
- Add cannot_audit() function non-Darwin systems that wraps auditon();
  required by OpenSSH BSM support.  Convert Darwin cannot_audit() into a
  function rather than a macro.
- Library build fixed on Darwin following include file tweaks.  The native
  Darwin sys/audit.h conflicts with bsm/audit.h due to duplicate types, so
  for now we force bsm_wrappers.c to not perform a nested include of
  sys/audit.h.

Obtained from: TrustedBSD Project

18 years agoAdd an OpenIPMI mostly compatible driver. This driver was developed
ambrisko [Fri, 10 Feb 2006 20:51:35 +0000 (20:51 +0000)]
Add an OpenIPMI mostly compatible driver.  This driver was developed
to work with ipmitools.  It works with other tools that have an OpenIPMI
driver interface.  The port will need to get updated to used this.
I have not implemented the IPMB mode yet so ioctl's for that don't
really do much otherwise it should work like the OpenIPMI version.
The ipmi.h definitions was derived from the ipmitool header file.
The bus attachments are done for smbios and pci/smbios.  Differences
in bus probe order for modules/static are delt with.  ACPI attachment
should be done.

This drivers registers with the watchdod(4) interface

Work to do:
     - BT interface
     - IPMB mode

This has been tested on Dell PE2850, PE2650 & PE850 with i386 & amd64
kernel.

I will link this into the build on next week.

Tom Rhodes, helped me with the man page.

Sponsored by:   IronPort Systems Inc.
Inspired from:  ipmitool & Linux

18 years ago- Fix attribute id of HIDBatteryPower
markus [Fri, 10 Feb 2006 19:54:17 +0000 (19:54 +0000)]
- Fix attribute id of HIDBatteryPower
- Fix two typos in comments

Approved by: emax

18 years agoUpdate for rev 0.9.16.16 hal:
sam [Fri, 10 Feb 2006 19:07:08 +0000 (19:07 +0000)]
Update for rev 0.9.16.16 hal:
o add dfs+radar hooks; DFS is presently disabled in the hal
o channel and mode handling changes
o various api changes
o be more aggressive about iq calibration settling so ap mode
  operation is better immediately after startup
o rfkill/rfsilent sysctl support
o tpc ack/cts sysctl support

MFC after: 2 weeks

18 years agoresolve merge conflicts
sam [Fri, 10 Feb 2006 19:01:02 +0000 (19:01 +0000)]
resolve merge conflicts

MFC after: 2 weeks

18 years agoThis commit was generated by cvs2svn to compensate for changes in r155511,
sam [Fri, 10 Feb 2006 18:57:38 +0000 (18:57 +0000)]
This commit was generated by cvs2svn to compensate for changes in r155511,
which included commits to RCS files with non-trunk default branches.

18 years agoVersion 0.9.16.16:
sam [Fri, 10 Feb 2006 18:57:38 +0000 (18:57 +0000)]
Version 0.9.16.16:

o new chip support
o new platforms: powerpc-be-elf, sparc64-be-elf, and alpha-elf
  (alpha is untested, others are known to work)
o many fixes and improvements

MFC after: 2 weeks

18 years agoCleaned the memory initialization up, moved some defines from the framebuffer
rink [Fri, 10 Feb 2006 18:48:22 +0000 (18:48 +0000)]
Cleaned the memory initialization up, moved some defines from the framebuffer
to an include file.

Reviewed by: imp
Approved by: imp (mentor)

18 years agoBump the MODULE_VERSION for HEAD, as the vlan(4) API is different in
emaste [Fri, 10 Feb 2006 18:38:33 +0000 (18:38 +0000)]
Bump the MODULE_VERSION for HEAD, as the vlan(4) API is different in
RELENG_6, and would require a lower version number.

Requested by: glebius
Approved by: rwatson (mentor)

18 years agoCorrectly set MNTK_MPSAFE flag from the lower vnode's mount rather than
jhb [Fri, 10 Feb 2006 18:06:49 +0000 (18:06 +0000)]
Correctly set MNTK_MPSAFE flag from the lower vnode's mount rather than
always turning it on along with any flags set in the lower mount.

Tested by: kris
Reviewed by: jeff
MFC after: 3 days

18 years agoThis commit was generated by cvs2svn to compensate for changes in r155506,
harti [Fri, 10 Feb 2006 16:49:49 +0000 (16:49 +0000)]
This commit was generated by cvs2svn to compensate for changes in r155506,
which included commits to RCS files with non-trunk default branches.

18 years agoVendor patch: Make the link traps more RFC-conform.
harti [Fri, 10 Feb 2006 16:49:49 +0000 (16:49 +0000)]
Vendor patch: Make the link traps more RFC-conform.

Submitted by: glebius

18 years agoCorrect references to ipnat(8).
brueffer [Fri, 10 Feb 2006 15:23:07 +0000 (15:23 +0000)]
Correct references to ipnat(8).

MFC after: 3 days

18 years agoTest before modifying p_sflag to avoid unconditionally cache line
davidxu [Fri, 10 Feb 2006 14:59:16 +0000 (14:59 +0000)]
Test before modifying p_sflag to avoid unconditionally cache line
ping-pong on SMP.

18 years agoCorrect the meaning of the NO_SHARED knob.
brueffer [Fri, 10 Feb 2006 14:19:07 +0000 (14:19 +0000)]
Correct the meaning of the NO_SHARED knob.

MFC after: 3 days

18 years agoAvoid frobbing IFF_UP at any cost (which is close to
yar [Fri, 10 Feb 2006 11:01:10 +0000 (11:01 +0000)]
Avoid frobbing IFF_UP at any cost (which is close to
zero in this case.)  A kernel driver has IFF_DRV_RUNNING
at its full disposal while IFF_UP may be toggled only by
humans or their daemonic deputies from the userland.

MFC after: 3 days

18 years agoCall thread_stopped in thr_exit to notify parent that the child process
davidxu [Fri, 10 Feb 2006 03:34:29 +0000 (03:34 +0000)]
Call thread_stopped in thr_exit to notify parent that the child process
is now fully stopped, this was already in kse_exit().

18 years agopad for future statistics
sam [Thu, 9 Feb 2006 22:37:33 +0000 (22:37 +0000)]
pad for future statistics

MFC after: 2 weeks

18 years agoMinor tx path cleanups:
sam [Thu, 9 Feb 2006 22:34:34 +0000 (22:34 +0000)]
Minor tx path cleanups:
o assume all data frames have been classified so there's no need
  to check if QoS is being used, just fetch the wme priority from
  the mbuf
o fix double counting of noack frames
o fix nearby comment

MFC after: 2 weeks

18 years agocorrect handling of mbuf allocation failure when replenishing the rx list
sam [Thu, 9 Feb 2006 22:29:23 +0000 (22:29 +0000)]
correct handling of mbuf allocation failure when replenishing the rx list
(leave a printf for the moment, need to make a debug msg)

Obtained from: atheros
MFC after: 2 weeks

18 years agoBeacon timer setup fixes:
sam [Thu, 9 Feb 2006 22:21:53 +0000 (22:21 +0000)]
Beacon timer setup fixes:
o pull nexttbtt forward in adhoc mode too
o resync beacon timers on joining a bss or ibss as the tstamp we
  collected while scanning is almost certainly out of date

Note we may need to refine the ibss mode check in ath_recv_mgmt.

Reviewed by: avatar, dyoung
Obtained from: atheros
MFC after: 2 weeks

18 years agoonly start the cab queue if there are frames to send
sam [Thu, 9 Feb 2006 22:14:11 +0000 (22:14 +0000)]
only start the cab queue if there are frames to send

MFC after: 2 weeks

18 years agodebug fixups: reduce noise msgs, report channel flags on reset failure,
sam [Thu, 9 Feb 2006 22:12:28 +0000 (22:12 +0000)]
debug fixups: reduce noise msgs, report channel flags on reset failure,
mark data+link fields in descriptor dumps

MFC after: 2 weeks

18 years agoAdd a MODULE_VERSION so that other modules (perhaps third-party) can
emaste [Thu, 9 Feb 2006 22:11:58 +0000 (22:11 +0000)]
Add a MODULE_VERSION so that other modules (perhaps third-party) can
depend on this one.

Approved by: rwatson (mentor)

18 years agoPhantom beacon miss workaround: track the tsf of the last received
sam [Thu, 9 Feb 2006 22:03:26 +0000 (22:03 +0000)]
Phantom beacon miss workaround: track the tsf of the last received
frame and if we get a beacon miss interrupt ignore it if we've received
a frame within the beacon miss interval.  This should never trigger
and the handling at the net80211 layer should likewise deal with this
but it doesn't hurt and can suppress extranous probe request frames.
Note that we can legtimately get a bmiss when under heavy load.

MFC after: 2 weeks

18 years agouse a private task queue thread
sam [Thu, 9 Feb 2006 21:48:51 +0000 (21:48 +0000)]
use a private task queue thread

MFC after: 2 weeks

18 years agoadd adhoc demo mode support
sam [Thu, 9 Feb 2006 21:42:53 +0000 (21:42 +0000)]
add adhoc demo mode support

MFC after: 2 weeks

18 years agomake regdomain sysctl r/w in case it's possible to do this in the future
sam [Thu, 9 Feb 2006 21:36:00 +0000 (21:36 +0000)]
make regdomain sysctl r/w in case it's possible to do this in the future

MFC after: 2 weeks

18 years agocleanup rate setup
sam [Thu, 9 Feb 2006 21:31:48 +0000 (21:31 +0000)]
cleanup rate setup

MFC after: 2 weeks

18 years agoSet the M_ZERO flag when calling uma_zalloc() to allocate a syncache entry.
qingli [Thu, 9 Feb 2006 21:29:02 +0000 (21:29 +0000)]
Set the M_ZERO flag when calling uma_zalloc() to allocate a syncache entry.

Reviewed by: andre, glebius
MFC after: 3 days

18 years agoadd tx99 hooks
sam [Thu, 9 Feb 2006 21:28:11 +0000 (21:28 +0000)]
add tx99 hooks

MFC after: 2 weeks

18 years agomove hal statistics to softc; the per-node stats are overkill, they're
sam [Thu, 9 Feb 2006 21:23:44 +0000 (21:23 +0000)]
move hal statistics to softc; the per-node stats are overkill, they're
only used when operating in station mode

MFC after: 2 weeks

18 years agolookup the protection tx rate index in the rate tables
sam [Thu, 9 Feb 2006 21:17:28 +0000 (21:17 +0000)]
lookup the protection tx rate index in the rate tables
instead of using a known value

MFC after: 2 weeks

18 years agohonor net80211 mcast tx rate
sam [Thu, 9 Feb 2006 21:15:36 +0000 (21:15 +0000)]
honor net80211 mcast tx rate

MFC after: 2 weeks

18 years agocraft unique names for tx q + buffer mtx's to help with interpreting ktr data
sam [Thu, 9 Feb 2006 21:09:26 +0000 (21:09 +0000)]
craft unique names for tx q + buffer mtx's to help with interpreting ktr data

MFC after: 2 weeks

18 years agoallow the size of tx+rx buffer pools to be tuned
sam [Thu, 9 Feb 2006 21:03:25 +0000 (21:03 +0000)]
allow the size of tx+rx buffer pools to be tuned

MFC after: 2 weeks

18 years agolower try count on mgt (and ctl) frames to avoid clogging the tx queue
sam [Thu, 9 Feb 2006 20:57:48 +0000 (20:57 +0000)]
lower try count on mgt (and ctl) frames to avoid clogging the tx queue
and loading the bss when operating in ap mode under load; adjust recognition
of multi-rate retry to match

MFC after: 2 weeks

18 years agoUnbreak Promise SATAII/150 controllers caused by the DMA dump changes.
sos [Thu, 9 Feb 2006 20:54:42 +0000 (20:54 +0000)]
Unbreak Promise SATAII/150 controllers caused by the DMA dump changes.

18 years agoUnbreak dump on legacy systems. Needs a more clean solution but that is
sos [Thu, 9 Feb 2006 20:53:32 +0000 (20:53 +0000)]
Unbreak dump on legacy systems. Needs a more clean solution but that is
more intrusive and can wait 'til after 6.1

18 years agomove mgt frame tx rate responsibility from the rate control modules
sam [Thu, 9 Feb 2006 20:49:55 +0000 (20:49 +0000)]
move mgt frame tx rate responsibility from the rate control modules
to the driver; this avoids redundant logic and will be necessary
for future additions

MFC after: 2 weeks

18 years agosync with latest code in madwifi
sam [Thu, 9 Feb 2006 20:40:28 +0000 (20:40 +0000)]
sync with latest code in madwifi

Obtained from: madwifi
MFC after: 2 weeks

18 years agoPick some low-hanging release notes fruit from cvs imports.
bmah [Thu, 9 Feb 2006 17:30:13 +0000 (17:30 +0000)]
Pick some low-hanging release notes fruit from cvs imports.

New release notes:  BIND 9.3.2 (+MFC), IPFilter 4.1.10, tzdata2005r (+MFC).

Deleted release notes:  libpcap and tcpdump updates (these updates
were actually a part of 6.0-RELEASE and therefore aren't "new"
anymore).

18 years agoActually fix the DHCP handling, by using more robust detection of DHCP
delphij [Thu, 9 Feb 2006 14:43:42 +0000 (14:43 +0000)]
Actually fix the DHCP handling, by using more robust detection of DHCP
setting.  While there, add some more debug messages explaining what is
happening, and some comments to improve readability.

Submitted by: ceri

18 years agoFix two important typos in watchdog handling:
glebius [Thu, 9 Feb 2006 12:57:17 +0000 (12:57 +0000)]
Fix two important typos in watchdog handling:

- Restart watchdog if we *did* processed any descriptors. [1]
- Log the watchdog event if the link is *up*. [2]

PR: kern/92948 [1]
Submitted by: Mihail Balikov <mihail.balikov interbgc.com> [1]
PR: kern/92895 [2]
Submitted by: Vladimir Ivanov <wawa yandex-team.ru> [2]

18 years agoRefuse to install invalid ARP entries.
glebius [Thu, 9 Feb 2006 12:49:39 +0000 (12:49 +0000)]
Refuse to install invalid ARP entries.

Submitted by: Vitaliy Ovsyannikov <V.Ovsyannikov kr.ru>

18 years agoMFi386: revision 1.622.
nyan [Thu, 9 Feb 2006 11:50:22 +0000 (11:50 +0000)]
MFi386: revision 1.622.

> Clear carry flag in get_mcontext so that setcontext does not
> return a bogus error.

MFC after: 3 days

18 years agoThe asr driver was moved to NOTES for i386. So 'nodevice asr' is not
nyan [Thu, 9 Feb 2006 11:43:13 +0000 (11:43 +0000)]
The asr driver was moved to NOTES for i386.  So 'nodevice asr' is not
needed anymore.

18 years ago- Increase maximum number of interfaces to 2048.
glebius [Thu, 9 Feb 2006 11:42:17 +0000 (11:42 +0000)]
- Increase maximum number of interfaces to 2048.
- Regroup softc so that frequently used elements are
  grouped in the beginning, while the interfaces
  array is at the end.

18 years agoAvoid calling CPUID function 0x02 if the CPU reports no support for
yar [Thu, 9 Feb 2006 09:10:54 +0000 (09:10 +0000)]
Avoid calling CPUID function 0x02 if the CPU reports no support for
it.  The former code used to hang older Intel CPUs by trying to get
non-existent TLB info 2^32 times.

Reduce code duplication around the calls to CPUID 0x02 by using
do-while loops.

PR: i386/92977
Tested by: cy

18 years agoRedo the previous fix by setting the UMA_ZONE_ZINIT bit in the syncache
qingli [Wed, 8 Feb 2006 23:32:57 +0000 (23:32 +0000)]
Redo the previous fix by setting the UMA_ZONE_ZINIT bit in the syncache
zone, eliminating the need to call bzero() after each syncache entry
allocation.

Suggested by: glebius
Reviewed by: andre
MFC after: 3 days

18 years agoCatch the case when a subdisk has no provider or no consumer
le [Wed, 8 Feb 2006 21:32:45 +0000 (21:32 +0000)]
Catch the case when a subdisk has no provider or no consumer
attached to it.

18 years agoreally truncate long ssid's when !verbose
sam [Wed, 8 Feb 2006 19:24:55 +0000 (19:24 +0000)]
really truncate long ssid's when !verbose

MFC after: 1 week

18 years agoset the mgt frame tx timer before dispatching the frame to the
sam [Wed, 8 Feb 2006 17:30:31 +0000 (17:30 +0000)]
set the mgt frame tx timer before dispatching the frame to the
driver; this closes a race where a response could be processed
before the timer was started and cause a RUN->SCAN state change
when operating in station mode

Reviewed by: avatar, dyoung
MFC after: 1 week

18 years agoThe bar.com is an existing domain. Instead of trying to listen on
glebius [Wed, 8 Feb 2006 11:55:03 +0000 (11:55 +0000)]
The bar.com is an existing domain. Instead of trying to listen on
64.15.205.248 address, change host to "foobar". This won't be
resolvable in most cases, so administrator will need to change it.
Also do not send our trap messages to 64.15.205.248, use localhost
as default traphost.

18 years agoFix a bug which puts an incorrect checksum of CHECKSUM.MD5
hrs [Wed, 8 Feb 2006 11:18:29 +0000 (11:18 +0000)]
Fix a bug which puts an incorrect checksum of CHECKSUM.MD5
into the CHECKSUM.MD5 file itself.

Approved by: re (scottl)
MFC after: 3 days

18 years agoSimplify system time accounting for profiling.
phk [Wed, 8 Feb 2006 08:09:17 +0000 (08:09 +0000)]
Simplify system time accounting for profiling.

Rename struct thread's td_sticks to td_pticks, we will need the
other name for more appropriately named use shortly.  Reduce it
from uint64_t to u_int.

Clear td_pticks whenever we enter the kernel instead of recording
its value as reference for userret().  Use the absolute value of
td->pticks in userret() and eliminate third argument.

18 years agoFix for an inappropriate bzero of the ICMPv6 stats. The code was zero'ing the wrong...
gnn [Wed, 8 Feb 2006 07:16:46 +0000 (07:16 +0000)]
Fix for an inappropriate bzero of the ICMPv6 stats.  The code was zero'ing the wrong structure member but setting the correct one.

Submitted by: James dot Juran at baesystems dot com
Reviewed by: gnn
MFC after: 1 week

18 years agoTeach gbde(8) to use a key file in addition to a passphrase. This
cperciva [Wed, 8 Feb 2006 06:52:15 +0000 (06:52 +0000)]
Teach gbde(8) to use a key file in addition to a passphrase.  This
makes it practical to use GBDE for "something you have plus something
you know" security together with a USB flash drive.

Reviewed by: phk
MFC after: 7 days

18 years agoGarbage collect a file that no longe is used (replaced by mpi_log_fc.h).
mjacob [Wed, 8 Feb 2006 05:58:25 +0000 (05:58 +0000)]
Garbage collect a file that no longe is used (replaced by mpi_log_fc.h).

18 years ago".Pp before .Sh is redundant, the latter asserts for a vertical space
marcel [Wed, 8 Feb 2006 00:11:41 +0000 (00:11 +0000)]
".Pp before .Sh is redundant, the latter asserts for a vertical space
 already."
-- ru@

18 years agoAcquire vnode lock around call to VOP_GETATTR() in audit_record_write().
rwatson [Tue, 7 Feb 2006 23:44:31 +0000 (23:44 +0000)]
Acquire vnode lock around call to VOP_GETATTR() in audit_record_write().
In the future, we may want to acquire the lock early in the function and
hold it across calls to vn_rdwr(), etc, to avoid multiple acquires.

Spotted by: kris (bugmagnet)
Obtained from: TrustedBSD Project

18 years agoSet the MAC address after we just read it at attach time, as it seems needed.
cognet [Tue, 7 Feb 2006 21:31:13 +0000 (21:31 +0000)]
Set the MAC address after we just read it at attach time, as it seems needed.

18 years agoModify the way we account for CPU time spent (step 1)
phk [Tue, 7 Feb 2006 21:22:02 +0000 (21:22 +0000)]
Modify the way we account for CPU time spent (step 1)

Keep track of time spent by the cpu in various contexts in units of
"cputicks" and scale to real-world microsec^H^H^H^H^H^H^H^Hclock_t
only when somebody wants to inspect the numbers.

For now "cputicks" are still derived from the current timecounter
and therefore things should by definition remain sensible also on
SMP machines.  (The main reason for this first milestone commit is
to verify that hypothesis.)

On slower machines, the avoided multiplications to normalize timestams
at every context switch, comes out as a 5-7% better score on the
unixbench/context1 microbenchmark.  On more modern hardware no change
in performance is seen.

18 years agoSet m_pkthdr.len and m_pkthdr.rcvif.
cognet [Tue, 7 Feb 2006 20:48:52 +0000 (20:48 +0000)]
Set m_pkthdr.len and m_pkthdr.rcvif.

18 years agoThe code in rn_walktree_from() that checks if we backed up too far
qingli [Tue, 7 Feb 2006 20:25:39 +0000 (20:25 +0000)]
The code in rn_walktree_from() that checks if we backed up too far
did not stop at the right node. Change the backtracking check from
smaller-than to smaller-or-equal to prevent this from happening.
While here fix one additional problem where the insertion of the
default route traversed the entire tree.

PR: kern/38752
Submitted by: qingli (before I became committer)
Reviewed by: andre
MFC after: 3 days

18 years ago- Add default choice option for --yesno box.
jkim [Tue, 7 Feb 2006 20:14:07 +0000 (20:14 +0000)]
- Add default choice option for --yesno box.
- Reduce trivial code duplications.

18 years agoRemove two unnecessary type casts, of which both had a typo in
qingli [Tue, 7 Feb 2006 20:09:02 +0000 (20:09 +0000)]
Remove two unnecessary type casts, of which both had a typo in
it anyways.

Approved by: andre
MFC after: 3 days

18 years agoFixes a crash due to the memory of the newly allocated syncache entry
qingli [Tue, 7 Feb 2006 19:59:46 +0000 (19:59 +0000)]
Fixes a crash due to the memory of the newly allocated syncache entry
in syncache_lookup() is not cleared and may lead to an arbitrary and
bogus rtentry pointer which later gets free'd.

Reviewed by: andre
MFC after: 3 days

18 years agoProvide some anti-footshooting. Don't allow the user to set the interval
jhb [Tue, 7 Feb 2006 18:59:47 +0000 (18:59 +0000)]
Provide some anti-footshooting.  Don't allow the user to set the interval
for acctwatch() runs to be negative or zero as this could result in either
a possible hang (or panic if INVARIANTS is on).  Previously the accounting
code handled the <= 0 case by calling acctwatch on every clock tick (eww!)
due to an implementation detail of callout_reset().  (Tick counts of
<= 0 are converted to 1).

MFC after: 3 days

18 years agoThis file isn't GENERIC, so change the header
imp [Tue, 7 Feb 2006 18:45:54 +0000 (18:45 +0000)]
This file isn't GENERIC, so change the header

18 years agoRather than pull in all the phy, just pull in the lxt phy that we need
imp [Tue, 7 Feb 2006 18:43:46 +0000 (18:43 +0000)]
Rather than pull in all the phy, just pull in the lxt phy that we need
for this board (although our lxtphy driver isn't claiming it, but that's
a different problem).

This saves 57k in this kernel.

18 years agoAllow one to subset phy. If you want the kitchen sink, use device
imp [Tue, 7 Feb 2006 18:41:56 +0000 (18:41 +0000)]
Allow one to subset phy.  If you want the kitchen sink, use device
miibus (like today).  If you want a subset, choose device mii and zero
or more phy to include.  We always include unkphy.  We make use of
the | functionality that ruslan recently added to config.

This allowed me to trim 57k from my KB9202 kernel.

18 years agoDetach the children before we delete them. This is a little cleaner
imp [Tue, 7 Feb 2006 18:38:51 +0000 (18:38 +0000)]
Detach the children before we delete them.  This is a little cleaner
than just deleting them.  Also add comments about why we do this.
Given the current behavior of delete_child, I don't think this changes
anything.  It just feels cleaner.

18 years agoINET doesn't belong to opt_inet6.h: INET6 belongs to it.
yar [Tue, 7 Feb 2006 18:13:56 +0000 (18:13 +0000)]
INET doesn't belong to opt_inet6.h: INET6 belongs to it.

18 years agoClean up some sysctl descriptions, debug messages etc.
brueffer [Tue, 7 Feb 2006 17:23:22 +0000 (17:23 +0000)]
Clean up some sysctl descriptions, debug messages etc.

Approved by: pjd
MFC after: 3 days

18 years ago- Add a kthread to periodically call acctwatch() when accounting is active
jhb [Tue, 7 Feb 2006 16:04:03 +0000 (16:04 +0000)]
- Add a kthread to periodically call acctwatch() when accounting is active
  instead of calling acctwatch() from softclock.  The acctwatch() function
  needs to hold an sx lock and also makes a VFS call, and neither of these
  are good things (or safe) to do from a callout.  The kthread only exists
  and is running when accounting is turned on; it is started and stopped
  as needed.  I didn't run acctwatch() via the thread taskqueue at Robert's
  request as he was worried that if the accounting file was over NFS the
  VFS_STAT() calls might stall other work on the taskqueue.
- Add an acct_disable() function to take care of closing the accounting
  vnode and cleaning up so we don't duplicate the same code in two
  different places.

MFC after: 3 days

18 years agoVendor patch: string_get_max() function to return strings with a maximum
harti [Tue, 7 Feb 2006 15:48:37 +0000 (15:48 +0000)]
Vendor patch: string_get_max() function to return strings with a maximum
SNMP string length.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r155429,
harti [Tue, 7 Feb 2006 15:48:37 +0000 (15:48 +0000)]
This commit was generated by cvs2svn to compensate for changes in r155429,
which included commits to RCS files with non-trunk default branches.

18 years agoFix queue drop logic when the queue overflows: decrement queue length.
rwatson [Tue, 7 Feb 2006 14:46:26 +0000 (14:46 +0000)]
Fix queue drop logic when the queue overflows: decrement queue length.

Obtained from: TrustedBSD Project

18 years agoRemove .depend when doing "make cleandir".
ru [Tue, 7 Feb 2006 13:37:26 +0000 (13:37 +0000)]
Remove .depend when doing "make cleandir".

18 years agoSince em(4) taskqueue is a new network context, we need to conditionally
glebius [Tue, 7 Feb 2006 13:11:13 +0000 (13:11 +0000)]
Since em(4) taskqueue is a new network context, we need to conditionally
lock Giant here.

Submitted by: Andrey V. Elsukov <bu7cher yandex.ru>

18 years agoFix five years old bug in ip_reass(): if we are using 'full' (i.e. including
oleg [Tue, 7 Feb 2006 11:48:10 +0000 (11:48 +0000)]
Fix five years old bug in ip_reass(): if we are using 'full' (i.e. including
pseudo header) hardware rx checksum offloading ip_reass() fails to calculate
TCP/UDP checksum for reassembled packet correctly.  This also should fix
recent 'NFS over UDP over bge' issue exposed by if_bge.c rev. 1.123

Reviewed by: sam (earlier version), bde
Approved by: glebius (mentor)
MFC after: 2 weeks

18 years agoRemove attempts to use mknod.
ceri [Tue, 7 Feb 2006 11:34:41 +0000 (11:34 +0000)]
Remove attempts to use mknod.

Approved by: jhb

18 years ago - No need to WANTPARENT when we're just going to vrele it in a deadlock
jeff [Tue, 7 Feb 2006 11:31:32 +0000 (11:31 +0000)]
 - No need to WANTPARENT when we're just going to vrele it in a deadlock
   prone way later.

Reported by: kkenn
MFC After: 3 days

18 years agoWhen there are no interesting information in output, exit with 0.
matteo [Tue, 7 Feb 2006 08:21:06 +0000 (08:21 +0000)]
When there are no interesting information in output, exit with 0.

PR: conf/92299
Submitted by: Petr Rehor <prehor@gmail.com>
Approved by: philip (mentor)
MFC after: 3 days

18 years agoThis commit was generated by cvs2svn to compensate for changes in r155420,
harti [Tue, 7 Feb 2006 07:58:11 +0000 (07:58 +0000)]
This commit was generated by cvs2svn to compensate for changes in r155420,
which included commits to RCS files with non-trunk default branches.

18 years agoVendor patch: remove the <CR>s that somehow crept in during initial commit.
harti [Tue, 7 Feb 2006 07:58:11 +0000 (07:58 +0000)]
Vendor patch: remove the <CR>s that somehow crept in during initial commit.