]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoThis commit was generated by cvs2svn to compensate for changes in r133594,
harti [Thu, 12 Aug 2004 19:17:29 +0000 (19:17 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133594,
which included commits to RCS files with non-trunk default branches.

19 years agoVendor patch: don't dump core when the config file cannot be opened.
harti [Thu, 12 Aug 2004 19:17:29 +0000 (19:17 +0000)]
Vendor patch: don't dump core when the config file cannot be opened.

Submitted by: Maxim Konovalov <maxim@macomnet.ru>

19 years agoNow that make more correctly handles variable assignments
harti [Thu, 12 Aug 2004 19:14:10 +0000 (19:14 +0000)]
Now that make more correctly handles variable assignments
in .MAKEFLAGS targets enable the regression test for this.

19 years agoWhen allocating the IPv6 header to stick in front of raw packet being
rwatson [Thu, 12 Aug 2004 18:31:36 +0000 (18:31 +0000)]
When allocating the IPv6 header to stick in front of raw packet being
sent via a raw IPv6 socket, use M_DONTWAIT not M_TRYWAIT, as we're
holding the raw pcb mutex.

Reported, tested by: kuriyama

19 years agoIn tcp6_ctlinput, lock tcbinfo around the call to syncache_unreach
dwmalone [Thu, 12 Aug 2004 18:19:36 +0000 (18:19 +0000)]
In tcp6_ctlinput, lock tcbinfo around the call to syncache_unreach
so that the locks held are the same as the IPv4 case.

Reviewed by: rwatson

19 years agoTrim trailing white space.
rwatson [Thu, 12 Aug 2004 18:06:21 +0000 (18:06 +0000)]
Trim trailing white space.

19 years ago- Introduce an ofw_bus kobj-interface for retrieving the OFW node and a
marius [Thu, 12 Aug 2004 17:41:33 +0000 (17:41 +0000)]
- Introduce an ofw_bus kobj-interface for retrieving the OFW node and a
  subset ("compatible", "device_type", "model" and "name") of the standard
  properties in drivers for devices on Open Firmware supported busses. The
  standard properties "reg", "interrupts" und "address" are not covered by
  this interface because they are only of interest in the respective bridge
  code. There's a remaining standard property "status" which is unclear how
  to support properly but which also isn't used in FreeBSD at present.
  This ofw_bus kobj-interface allows to replace the various (ebus_get_node(),
  ofw_pci_get_node(), etc.) and partially inconsistent (central_get_type()
  vs. sbus_get_device_type(), etc.) existing IVAR ones with a common one.
  This in turn allows to simplify and remove code-duplication in drivers for
  devices that can hang off of more than one OFW supported bus.
- Convert the sparc64 Central, EBus, FHC, PCI and SBus bus drivers and the
  drivers for their children to use the ofw_bus kobj-interface. The IVAR-
  interfaces of the Central, EBus and FHC are entirely replaced by this. The
  PCI bus driver used its own kobj-interface and now also uses the ofw_bus
  one. The IVARs special to the SBus, e.g. for retrieving the burst size,
  remain.
  Beware: this causes an ABI-breakage for modules of drivers which used the
  IVAR-interfaces, i.e. esp(4), hme(4), isp(4) and uart(4), which need to be
  recompiled.
  The style-inconsistencies introduced in some of the bus drivers will be
  fixed by tmm@ in a generic clean-up of the respective drivers later (he
  requested to add the changes in the "new" style).
- Convert the powerpc MacIO bus driver and the drivers for its children to
  use the ofw_bus kobj-interface. This invloves removing the IVARs related
  to the "reg" property which were unused and a leftover from the NetBSD
  origini of the code. There's no ABI-breakage caused by this because none
  of these driver are currently built as modules.
  There are other powerpc bus drivers which can be converted to the ofw_bus
  kobj-interface, e.g. the PCI bus driver, which should be done together
  with converting powerpc to use the OFW PCI code from sparc64.
- Make the SBus and FHC front-end of zs(4) and the sparc64 eeprom(4) take
  advantage of the ofw_bus kobj-interface and simplify them a bit.

Reviewed by: grehan, tmm
Approved by: re (scottl)
Discussed with: tmm
Tested with: Sun AX1105, AXe, Ultra 2, Ultra 60; PPC cross-build on i386

19 years agoMinor formatting fixes for lines > 80 characters
imp [Thu, 12 Aug 2004 17:26:22 +0000 (17:26 +0000)]
Minor formatting fixes for lines > 80 characters

19 years agoRe-delete the comment from r1.352.
green [Thu, 12 Aug 2004 17:22:28 +0000 (17:22 +0000)]
Re-delete the comment from r1.352.

19 years agoSkip dependent functions when finding the resource from _PRS to use for
njl [Thu, 12 Aug 2004 17:06:05 +0000 (17:06 +0000)]
Skip dependent functions when finding the resource from _PRS to use for
later calls to _SRS.  Add note that this code should be centralized at
some point.

Bug from: Jiawei Ye <leafy7382_AT_gmail.com>

19 years agoOnly print the link name if there is a link. For the hardwired case, don't
njl [Thu, 12 Aug 2004 17:04:19 +0000 (17:04 +0000)]
Only print the link name if there is a link.  For the hardwired case, don't
bother printing it.  This fixes a panic and acpi_name() has been more robust
as well.

Bug from: Tai-hwa Liang <avatar-at-mmlab.cse.yzu.edu.tw>

19 years agoAllow null handles to be passed into acpi_name().
njl [Thu, 12 Aug 2004 17:02:53 +0000 (17:02 +0000)]
Allow null handles to be passed into acpi_name().

19 years agoThis commit was generated by cvs2svn to compensate for changes in r133582,
kan [Thu, 12 Aug 2004 16:41:42 +0000 (16:41 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133582,
which included commits to RCS files with non-trunk default branches.

19 years agoConsolidate libiberty files in one place and avoid listing some
kan [Thu, 12 Aug 2004 16:36:52 +0000 (16:36 +0000)]
Consolidate libiberty files in one place and avoid listing some
of them in several places.

Noticed by: bsdimp

19 years agoAllow the ATM call control module to be built into the kernel.
harti [Thu, 12 Aug 2004 15:01:59 +0000 (15:01 +0000)]
Allow the ATM call control module to be built into the kernel.

19 years agoAdd the module build stuff for the ATM call control module.
harti [Thu, 12 Aug 2004 14:58:46 +0000 (14:58 +0000)]
Add the module build stuff for the ATM call control module.

19 years agoThis is the netgraph node framework for the user side call control
harti [Thu, 12 Aug 2004 14:22:00 +0000 (14:22 +0000)]
This is the netgraph node framework for the user side call control
node for ATM. This node implements the API to the signalling services.

19 years agoLoopback fix from Mathieu Sauve-Frankel:
mlaier [Thu, 12 Aug 2004 14:15:42 +0000 (14:15 +0000)]
Loopback fix from Mathieu Sauve-Frankel:
Add missing check for NULL in DIOCCHANGERULE. This prevents a crash
in certain rare cases.

19 years agoLoopback fix from Daniel Hartmeier:
mlaier [Thu, 12 Aug 2004 13:59:44 +0000 (13:59 +0000)]
Loopback fix from Daniel Hartmeier:
pf_cksum_fixup() was called without last argument from
normalization, also fixup checksum when random-id modifies ip_id.
This would previously lead to incorrect checksums for packets
modified by scrub random-id.

(Originally) Submitted by: yongari

19 years agoLoopback fix from Henning Brauer:
mlaier [Thu, 12 Aug 2004 13:54:50 +0000 (13:54 +0000)]
Loopback fix from Henning Brauer:
skip over interface addresses without IFA_ROUTE, fixes some issue
with pppd

PR: misc/69954

19 years agoThe geom(8) utility needs dynamic linker functionality to work, so it can't
pjd [Thu, 12 Aug 2004 13:15:52 +0000 (13:15 +0000)]
The geom(8) utility needs dynamic linker functionality to work, so it can't
be staticaly linked.
This fixes problems on systems compiled with NO_DYNAMICROOT.

19 years agoclarify that both ECC and non-ECC DIMMS work on 164SX.
wilko [Thu, 12 Aug 2004 12:42:03 +0000 (12:42 +0000)]
clarify that both ECC and non-ECC DIMMS work on 164SX.

Suggested by: ru

19 years agoSort in dictionary order.
tjr [Thu, 12 Aug 2004 12:36:04 +0000 (12:36 +0000)]
Sort in dictionary order.

Suggested by: ru

19 years agoFix example.
tjr [Thu, 12 Aug 2004 12:32:14 +0000 (12:32 +0000)]
Fix example.

19 years agoAdd support for the examination and modification of the devices.
harti [Thu, 12 Aug 2004 12:31:43 +0000 (12:31 +0000)]
Add support for the examination and modification of the devices.
This is implemented through SNMP and requires the ilmi daemon to
run on the system. To prevent bloat in rescue the atmconfig for
rescue is compiled without this stuff.

19 years agoImplement wcwidth() as an inline function.
tjr [Thu, 12 Aug 2004 12:19:11 +0000 (12:19 +0000)]
Implement wcwidth() as an inline function.

19 years agoAs long as we have a knob to force system scope threads, why not have
deischen [Thu, 12 Aug 2004 12:12:12 +0000 (12:12 +0000)]
As long as we have a knob to force system scope threads, why not have
a knob to force process scope threads.  If the environment variable
LIBPTHREAD_PROCESS_SCOPE is set, force all threads to be process
scope threads regardless of how the application creates them.  If
LIBPTHREAD_SYSTEM_SCOPE is set (forcing system scope threads), it
overrides LIBPTHREAD_PROCESS_SCOPE.

        $ # To force system scope threads
        $ LIBPTHREAD_SYSTEM_SCOPE=anything threaded_app
        $ # To force process scope threads
        $ LIBPTHREAD_PROCESS_SCOPE=anything threaded_app

19 years agoPut variable assignments on .MAKEFLAGS and .MFLAGS targets into
harti [Thu, 12 Aug 2004 11:49:55 +0000 (11:49 +0000)]
Put variable assignments on .MAKEFLAGS and .MFLAGS targets into
the .MAKEFLAGS variable so that these are also passed to sub-makes.
This makes the handling of variables in the command environment more
consistent.

PR: bin/68853
Submitted by: Martin Kamerhofer <data@sbox.tugraz.at>

19 years agoDocument incorrect handling of multibyte characters.
tjr [Thu, 12 Aug 2004 11:34:34 +0000 (11:34 +0000)]
Document incorrect handling of multibyte characters.

19 years agoInclude _ctype.h instead of ctype.h to avoid namespace pollution.
tjr [Thu, 12 Aug 2004 10:29:14 +0000 (10:29 +0000)]
Include _ctype.h instead of ctype.h to avoid namespace pollution.

19 years agoMove some internal macros and inlines from ctype.h to a new file, _ctype.h,
tjr [Thu, 12 Aug 2004 09:33:47 +0000 (09:33 +0000)]
Move some internal macros and inlines from ctype.h to a new file, _ctype.h,
which has been repo-copied from ctype.h. This will allow us to remove
namespace pollution from <wctype.h> and to make wcwidth() an inline function
without introducing more pollution.

19 years agoChange RU for articles.ent and books.ent back to EN for consistency
den [Thu, 12 Aug 2004 08:44:52 +0000 (08:44 +0000)]
Change RU for articles.ent and books.ent back to EN for consistency
This modification was not really needed.

Noticed by: hrs
Obtained from: The FreeBSD Russian Documentation Project

19 years agoFix two cases of incorrect IPQ_UNLOCK'ing in the merged ip_reass() function.
andre [Thu, 12 Aug 2004 08:37:42 +0000 (08:37 +0000)]
Fix two cases of incorrect IPQ_UNLOCK'ing in the merged ip_reass() function.
The first one was going to 'dropfrag', which unlocks the IPQ, before the lock
was aquired; The second one doing a unlock and then a 'goto dropfrag' which
led to a double-unlock.

Tripped over by: des

19 years agoChange the order of ata_dmainit/ata_allocate in preparation of
sos [Thu, 12 Aug 2004 08:20:36 +0000 (08:20 +0000)]
Change the order of ata_dmainit/ata_allocate in preparation of
supporting new chipsets where this is needed.

19 years ago - Introduce a new flag KEF_HOLD that prevents sched_add() from doing a
jeff [Thu, 12 Aug 2004 07:56:33 +0000 (07:56 +0000)]
 - Introduce a new flag KEF_HOLD that prevents sched_add() from doing a
   migration.  Use this in sched_prio() and sched_switch() to stop us from
   migrating threads that are in short term sleeps or are runnable.  These
   extra migrations were added in the patches to support KSE.
 - Only set NEEDRESCHED if the thread we're adding in sched_add() is a
   lower priority and is being placed on the current queue.
 - Fix some minor whitespace problems.

19 years agoJoin the Russian committers, and only use one true country of birth.
ru [Thu, 12 Aug 2004 06:50:44 +0000 (06:50 +0000)]
Join the Russian committers, and only use one true country of birth.
Sorry for the churn.  ;)

19 years agoMove towards isa attachment for pccbb. This is a work in progress, but
imp [Thu, 12 Aug 2004 06:50:29 +0000 (06:50 +0000)]
Move towards isa attachment for pccbb.  This is a work in progress, but
works well with the pci attachment.

19 years agoChange version string to distinguish our modified version of GNU grep
tjr [Thu, 12 Aug 2004 06:45:31 +0000 (06:45 +0000)]
Change version string to distinguish our modified version of GNU grep
from the original.

19 years agoUpdate version strings to better reflect reality.
tjr [Thu, 12 Aug 2004 06:40:31 +0000 (06:40 +0000)]
Update version strings to better reflect reality.

19 years agoSpell check the 20040806 entry.
ru [Thu, 12 Aug 2004 06:25:05 +0000 (06:25 +0000)]
Spell check the 20040806 entry.

19 years agoHook xalloc-die.c up to the build.
tjr [Thu, 12 Aug 2004 06:02:28 +0000 (06:02 +0000)]
Hook xalloc-die.c up to the build.

19 years agoThis commit was generated by cvs2svn to compensate for changes in r133546,
tjr [Thu, 12 Aug 2004 05:59:33 +0000 (05:59 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133546,
which included commits to RCS files with non-trunk default branches.

19 years agoImport of GNU sort from coreutils CVS (trimmed)
tjr [Thu, 12 Aug 2004 05:59:33 +0000 (05:59 +0000)]
Import of GNU sort from coreutils CVS (trimmed)

19 years agoMerge local changes.
tjr [Thu, 12 Aug 2004 05:46:04 +0000 (05:46 +0000)]
Merge local changes.

19 years agoImport of GNU sort from coreutils CVS (trimmed)
tjr [Thu, 12 Aug 2004 05:37:46 +0000 (05:37 +0000)]
Import of GNU sort from coreutils CVS (trimmed)

19 years agoThis commit was generated by cvs2svn to compensate for changes in r133543,
tjr [Thu, 12 Aug 2004 05:37:46 +0000 (05:37 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133543,
which included commits to RCS files with non-trunk default branches.

19 years agoDestroy mutexes on detach.
scottl [Thu, 12 Aug 2004 05:31:17 +0000 (05:31 +0000)]
Destroy mutexes on detach.

19 years agoChange FREE() to free()
scottl [Thu, 12 Aug 2004 05:09:17 +0000 (05:09 +0000)]
Change FREE() to free()

19 years agoRemove the AAC_LOCK macros. They no longer abstract anything and only
scottl [Thu, 12 Aug 2004 05:05:06 +0000 (05:05 +0000)]
Remove the AAC_LOCK macros.  They no longer abstract anything and only
obfuscate the code.  No functional differences.

19 years agoRelease the sync fib after the controller has been shut down. This also
scottl [Thu, 12 Aug 2004 04:47:32 +0000 (04:47 +0000)]
Release the sync fib after the controller has been shut down.  This also
releases the I/O lock instead of just leaking it.

19 years agoSet IFF_NEEDSGIANT for fwe network interface since the firewire
rwatson [Thu, 12 Aug 2004 03:02:16 +0000 (03:02 +0000)]
Set IFF_NEEDSGIANT for fwe network interface since the firewire
framework isn't yet MPSAFE.

Approved by: simokawa

19 years agoBack in the USSR (for me)
ache [Thu, 12 Aug 2004 02:23:24 +0000 (02:23 +0000)]
Back in the USSR (for me)

19 years agoThis commit was generated by cvs2svn to compensate for changes in r133534,
kan [Thu, 12 Aug 2004 02:21:58 +0000 (02:21 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133534,
which included commits to RCS files with non-trunk default branches.

19 years agoImport one more file for Gcc 3.4.2 20040728.
kan [Thu, 12 Aug 2004 02:21:58 +0000 (02:21 +0000)]
Import one more file for Gcc 3.4.2 20040728.

19 years agoFix the PRT entry code in acpi_pci_link to always add the entry, even if
njl [Thu, 12 Aug 2004 02:06:19 +0000 (02:06 +0000)]
Fix the PRT entry code in acpi_pci_link to always add the entry, even if
there is no irq link.  Since we now use the stored copy of PRT, not the
one that used to be passed into acpi_pcib_route_interrupt(), we need it in
the list. [1]

Fix a bug in acpi_pci_find_prt() where we weren't checking the bus, thus
choosing the wrong PRT entry to use for routing the link.  Also, add a
printf for the case where the PRT entry is not found as this should not
happen.

Tested by: marcel [1]

19 years agoWhen udp_send() fails, make sure to free the control mbufs as well as
rwatson [Thu, 12 Aug 2004 01:34:27 +0000 (01:34 +0000)]
When udp_send() fails, make sure to free the control mbufs as well as
the data mbuf.  This was done in most error cases, but not the case
where the inpcb pointer is surprisingly NULL.

19 years agoadd comment about needing to upgrade loader.rc on i386 systems that have
jmg [Thu, 12 Aug 2004 00:00:15 +0000 (00:00 +0000)]
add comment about needing to upgrade loader.rc on i386 systems that have
problems loading modules...

Solution by: ru

19 years agoMFp4: Simplify code a bit:
pjd [Wed, 11 Aug 2004 23:41:53 +0000 (23:41 +0000)]
MFp4: Simplify code a bit:
- Remove kern.geom.mirror.sync_block_size sysctl. It is quite obvious that we
  want to use the biggest size possible.
- Do not use UMA zone for sync data allocations. There could be only one
  synchronization request per synchronized disk at a time, so allocate memory
  for one request on whole synchronization process related to one disk.

Tested by synchronizing one component (out of three) and by synchronizing
two components (out of three) in parallel.

19 years agoMark end of frames.
davidxu [Wed, 11 Aug 2004 23:23:05 +0000 (23:23 +0000)]
Mark end of frames.

19 years agoActually, HARDCODED flag isn't stored in metadata, so don't bother
pjd [Wed, 11 Aug 2004 22:16:42 +0000 (22:16 +0000)]
Actually, HARDCODED flag isn't stored in metadata, so don't bother
dumping it.

19 years ago- Fix typo.
pjd [Wed, 11 Aug 2004 22:12:44 +0000 (22:12 +0000)]
- Fix typo.
- Dump HARDCODED flag.

19 years agoBe explicit that a driver may assume softc is preserved when returning 0
imp [Wed, 11 Aug 2004 21:52:31 +0000 (21:52 +0000)]
Be explicit that a driver may assume softc is preserved when returning 0

19 years agoAdd more aliases for common CPUTYPE spellings
obrien [Wed, 11 Aug 2004 21:51:27 +0000 (21:51 +0000)]
Add more aliases for common CPUTYPE spellings

19 years agoDeorbit the fxp tuning hint (hint.fxp.UNIT_NUMBER.ipcbxmit_disable)
truckman [Wed, 11 Aug 2004 21:41:25 +0000 (21:41 +0000)]
Deorbit the fxp tuning hint (hint.fxp.UNIT_NUMBER.ipcbxmit_disable)
introduced in if_fxp.c revision 1.180.  The bug fix committed in
revision 1.180 fixed the packet truncation problem.

19 years agoFix the country of birth.
ru [Wed, 11 Aug 2004 21:23:42 +0000 (21:23 +0000)]
Fix the country of birth.

Requested by: dima

19 years agoUnbreak after struct resource was hidden. Tested with EBus front-end on
marius [Wed, 11 Aug 2004 21:19:31 +0000 (21:19 +0000)]
Unbreak after struct resource was hidden. Tested with EBus front-end on
Sun AXe board.

19 years ago- Use the rman_get_* functions instead of reaching into struct resource.
marius [Wed, 11 Aug 2004 21:09:40 +0000 (21:09 +0000)]
- Use the rman_get_* functions instead of reaching into struct resource.
- Remove __RMAN_RESORUCE_VISIBLE again. It's no longer required either
  because of the above change or because struct rman is no longer hidden.

Reviewed by: grehan
Tested by: cross-compile on i386

19 years agoProperly keep track of how many kses are on the system run queue(s).
julian [Wed, 11 Aug 2004 20:54:48 +0000 (20:54 +0000)]
Properly keep track of how many kses are on the system run queue(s).

19 years agoAdd some more verbose warning/error messages to help with users reporting
njl [Wed, 11 Aug 2004 20:37:24 +0000 (20:37 +0000)]
Add some more verbose warning/error messages to help with users reporting
problems with irq routing.

19 years agoCorrectly export the size of our softc to newbus in our EISA attachement.
gibbs [Wed, 11 Aug 2004 20:35:55 +0000 (20:35 +0000)]
Correctly export the size of our softc to newbus in our EISA attachement.
This avoids a panic upon first softc field reference.

19 years agoBackout removal of UMA_ZONE_NOFREE flag for all zones which are established
andre [Wed, 11 Aug 2004 20:30:08 +0000 (20:30 +0000)]
Backout removal of UMA_ZONE_NOFREE flag for all zones which are established
for structures with timers in them.  It might be that a timer might fire
even when the associated structure has already been free'd.  Having type-
stable storage in this case is beneficial for graceful failure handling and
debugging.

Discussed with: bosko, tegge, rwatson

19 years agoThe last commit had one too many libs.
obrien [Wed, 11 Aug 2004 19:19:26 +0000 (19:19 +0000)]
The last commit had one too many libs.

19 years agoConvert the routing table to use an UMA zone for rtentries. The zone is
andre [Wed, 11 Aug 2004 17:26:56 +0000 (17:26 +0000)]
Convert the routing table to use an UMA zone for rtentries.  The zone is
called "rtentry".

This saves a considerable amount of kernel memory.  R_Zmalloc previously
used 256 byte blocks (plus kmalloc overhead) whereas UMA only needs 132
bytes.

Idea from: OpenBSD

19 years agoRemove badly broken pcic driver for NEWCARD
imp [Wed, 11 Aug 2004 17:24:42 +0000 (17:24 +0000)]
Remove badly broken pcic driver for NEWCARD

19 years agoRemove pcic for NEWCARD
imp [Wed, 11 Aug 2004 17:23:32 +0000 (17:23 +0000)]
Remove pcic for NEWCARD

19 years agoRemove references to pcic for newcard from NOTES.
imp [Wed, 11 Aug 2004 17:22:37 +0000 (17:22 +0000)]
Remove references to pcic for newcard from NOTES.

19 years agoRemove the UMA_ZONE_NOFREE flag to all uma_zcreate() calls in the IP and
andre [Wed, 11 Aug 2004 17:08:31 +0000 (17:08 +0000)]
Remove the UMA_ZONE_NOFREE flag to all uma_zcreate() calls in the IP and
TCP code.  This flag would have prevented giving back excessive free slabs
to the global pool after a transient peak usage.

19 years agoLink to the build again. The problem was that we need a new
harti [Wed, 11 Aug 2004 16:47:12 +0000 (16:47 +0000)]
Link to the build again. The problem was that we need a new
gensnmptree for the .def file to parse. This was fixed in
Makefile.inc1.

19 years agoEnsure that a new gensnmptree is built. This should fix the problems
harti [Wed, 11 Aug 2004 16:45:20 +0000 (16:45 +0000)]
Ensure that a new gensnmptree is built. This should fix the problems
with the snmp_atm module.

19 years agoAdd my birthday.
marck [Wed, 11 Aug 2004 16:18:18 +0000 (16:18 +0000)]
Add my birthday.

Approved by: den (mentor)

19 years agoAgain something bad is happening so comment out the ATM module.
harti [Wed, 11 Aug 2004 16:01:46 +0000 (16:01 +0000)]
Again something bad is happening so comment out the ATM module.

19 years agotypo
eik [Wed, 11 Aug 2004 15:47:13 +0000 (15:47 +0000)]
typo

Approved by: ru

19 years agoFWIIW, make the `test' target in pkg_install/version actually work.
eik [Wed, 11 Aug 2004 15:46:10 +0000 (15:46 +0000)]
FWIIW, make the `test' target in pkg_install/version actually work.

19 years agoRe-work ACPI PCI IRQ routing (_PRT, link devices). The old approach was
njl [Wed, 11 Aug 2004 14:52:50 +0000 (14:52 +0000)]
Re-work ACPI PCI IRQ routing (_PRT, link devices).  The old approach was
incomplete in that the PRT routing was not aware of link programming.
Fix this by doing all routing through the link devices.  The new algorithm
for setting up links is:

1. Read _CRS to get current setting.  If invalid (not in _PRS), then set
   to 0.
2. Attempt to call _DIS on the link.  If successful, mark the link as not
   routed.  Otherwise, assume it still is.

Then when a routing request occurs:

3. Update weights for all IRQs
4. Attempt to route the initial IRQ if valid
5. If that fails, walk through the sorted list, attempting to route IRQs.
6. Configure the trigger/polarity based on _PRS.

Other changes:
* Add acpi_pci_find_prt() to look up the PRT entry for a given device and
  acpi_pci_link_route() to select/route the best IRQ for it.
* Remove duplicated code in acpi_pcib_route_interrupt() that picked the
  first IRQ from _PRS.
* Remove unneeded arguments from acpi_pcib_resume() and friends.
* Ignore _STA on link devices but report if it seems strange.
* Add a prt_source handle to the PRT structure since the ACPI struct
  ACPI_PCI_ROUTING_TABLE uses a fixed-size entry for it.  We'll need to
  dynamically size this object if we want to use it the same way ACPI-CA
  does.  Null-terminate the source.

Tested by: Luo Hong <luohong99_at_mails.tsinghua.edu.cn>,
Jeffrey Katcher <jmkatcher_at_yahoo.com>
Info from: jhb, Len Brown (Intel)

19 years agofix description of the macros for BUS_DMASYNC opertions. It's from the CPU's
jmg [Wed, 11 Aug 2004 14:52:33 +0000 (14:52 +0000)]
fix description of the macros for BUS_DMASYNC opertions.  It's from the CPU's
point of view...  Add a reference to a better description of what the ops
do...

Reviewed by: scottl

19 years agoIncrease default kern.geom.stripe.maxmem to 50 elements.
pjd [Wed, 11 Aug 2004 12:57:17 +0000 (12:57 +0000)]
Increase default kern.geom.stripe.maxmem to 50 elements.

19 years agoMake use of in_localip() function and replace previous direct LIST_FOREACH
andre [Wed, 11 Aug 2004 12:32:10 +0000 (12:32 +0000)]
Make use of in_localip() function and replace previous direct LIST_FOREACH
loops over INADDR_HASH.

19 years agoNow, after the import of NgATM-1.1, re-enable the atm module.
harti [Wed, 11 Aug 2004 12:27:31 +0000 (12:27 +0000)]
Now, after the import of NgATM-1.1, re-enable the atm module.

19 years agoExclusion list for easier import.
harti [Wed, 11 Aug 2004 12:25:09 +0000 (12:25 +0000)]
Exclusion list for easier import.

19 years agoThis commit was generated by cvs2svn to compensate for changes in r133492,
harti [Wed, 11 Aug 2004 12:21:36 +0000 (12:21 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133492,
which included commits to RCS files with non-trunk default branches.

19 years agoVirgin import of NgATM shared kernel/user part 1.1
harti [Wed, 11 Aug 2004 12:21:36 +0000 (12:21 +0000)]
Virgin import of NgATM shared kernel/user part 1.1

19 years agoExclusion list for easier import.
harti [Wed, 11 Aug 2004 12:19:32 +0000 (12:19 +0000)]
Exclusion list for easier import.

19 years agoThis commit was generated by cvs2svn to compensate for changes in r133488,
harti [Wed, 11 Aug 2004 12:17:54 +0000 (12:17 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133488,
which included commits to RCS files with non-trunk default branches.

19 years agoVirgin import of NgATM user part 1.1
harti [Wed, 11 Aug 2004 12:17:54 +0000 (12:17 +0000)]
Virgin import of NgATM user part 1.1

19 years agoWhen sending request once again because of ENOMEM, reset bio_children
pjd [Wed, 11 Aug 2004 12:04:35 +0000 (12:04 +0000)]
When sending request once again because of ENOMEM, reset bio_children
and bio_inbed fields to 0. Without this change we can end up with
I/O leakage in some rare situations.
I tested this change by putting failure probability mechanism simlar
to this used in NOP class into g_clone_bio(9) function, so it was
able to return NULL with the given probability.

Discussed with: phk

19 years agoAdd the function in_localip() which returns 1 if an internet address is for
andre [Wed, 11 Aug 2004 11:49:48 +0000 (11:49 +0000)]
Add the function in_localip() which returns 1 if an internet address is for
the local host and configured on one of its interfaces.

19 years agoOnly invoke verify_path() for verrevpath and versrcreach when we have an IP packet.
andre [Wed, 11 Aug 2004 11:41:11 +0000 (11:41 +0000)]
Only invoke verify_path() for verrevpath and versrcreach when we have an IP packet.

19 years agoTry harder to not panic on 'stop -f'.
pjd [Wed, 11 Aug 2004 11:10:46 +0000 (11:10 +0000)]
Try harder to not panic on 'stop -f'.
After the commit, this command should be really safe to use.

19 years agoEnforce the check that MAKEOBJDIRPREFIX (if set) is set in environment
ru [Wed, 11 Aug 2004 11:02:26 +0000 (11:02 +0000)]
Enforce the check that MAKEOBJDIRPREFIX (if set) is set in environment
only, and not as a global (in /etc/make.conf) or command-line variable.

MAKEOBJDIRPREFIX has never been a global or command-line variable, and
the fact that it works in some scenarios for "make buildworld" doesn't
make it any more correct.  Using it as a global or command-line variable
is error prone, discouraged, costs us lot of false build reports, etc.

This commit is aimed to fix it once and for all.

Anyone potentially objecting to this change is encouraged to read the
make(1) and make.conf(5) manpages, and the comments regarding the use
of the MAKEOBJDIRPREFIX variable in /usr/share/mk/bsd.obj.mk and
/usr/share/examples/etc/make.conf.

19 years agoOnly check for local broadcast addresses if the mbuf is flagged with M_BCAST.
andre [Wed, 11 Aug 2004 10:49:56 +0000 (10:49 +0000)]
Only check for local broadcast addresses if the mbuf is flagged with M_BCAST.