]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoThis commit was generated by cvs2svn to compensate for changes in r151513,
Bill Paul [Thu, 20 Oct 2005 16:31:54 +0000 (16:31 +0000)]
This commit was generated by cvs2svn to compensate for changes in r151513,
which included commits to RCS files with non-trunk default branches.

18 years agoImport a fixed version of driver_ndis.c from the vendor. The existing
Bill Paul [Thu, 20 Oct 2005 16:31:54 +0000 (16:31 +0000)]
Import a fixed version of driver_ndis.c from the vendor. The existing
version has a bug where it fails to properly cancel the polling loop
that periodically queries the BSSID (this is done to detect the
association/disassociation state). The timeout is supposed to fire
once a second, but the eloop_cancel_timeout() call uses a different
'user data' value than what was passed to eloop_register_timeout(),
so cancelling the timeouts fails. This results in an additional timeout
being created each time an EAPOL packet is received, which can lead
to dozens of unwanted timeouts firing every second instead of just one.

18 years agoAdd entry for the spin mutex used by the hptmv(4) driver.
John Baldwin [Thu, 20 Oct 2005 14:49:59 +0000 (14:49 +0000)]
Add entry for the spin mutex used by the hptmv(4) driver.

MFC after:  1 day
Tested by: Philip Kizer pckizer at nostrum dot com

18 years agoForce an upgrade to Groff 1.19.2.
Ruslan Ermilov [Thu, 20 Oct 2005 11:22:30 +0000 (11:22 +0000)]
Force an upgrade to Groff 1.19.2.

18 years agoRemove a diagnostic message that can't occur: we lost the ability to
Diomidis Spinellis [Thu, 20 Oct 2005 11:20:55 +0000 (11:20 +0000)]
Remove a diagnostic message that can't occur: we lost the ability to
handle the old filesystem format on 2002/06/21.

18 years agoUpdate upgrade instructions.
Ruslan Ermilov [Thu, 20 Oct 2005 11:18:24 +0000 (11:18 +0000)]
Update upgrade instructions.

18 years agoUse an ihandle_t to store the instance handle of an opened device
Marius Strobl [Thu, 20 Oct 2005 11:14:34 +0000 (11:14 +0000)]
Use an ihandle_t to store the instance handle of an opened device
instead of a phandle_t (package handle). Since both are typedefed
to unsigned int, this is more or less cosmetic.

18 years agoRemove some conversion leftovers.
Ruslan Ermilov [Thu, 20 Oct 2005 11:10:43 +0000 (11:10 +0000)]
Remove some conversion leftovers.

18 years agoUpgrade to Groff 1.19.2.
Ruslan Ermilov [Thu, 20 Oct 2005 10:57:53 +0000 (10:57 +0000)]
Upgrade to Groff 1.19.2.

18 years agoMerge Groff 1.19.2 changes.
Ruslan Ermilov [Thu, 20 Oct 2005 10:53:15 +0000 (10:53 +0000)]
Merge Groff 1.19.2 changes.

18 years agoUse stock (FSF) version of this file.
Ruslan Ermilov [Thu, 20 Oct 2005 10:52:34 +0000 (10:52 +0000)]
Use stock (FSF) version of this file.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r151500,
Ruslan Ermilov [Thu, 20 Oct 2005 10:47:26 +0000 (10:47 +0000)]
This commit was generated by cvs2svn to compensate for changes in r151500,
which included commits to RCS files with non-trunk default branches.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r151497,
Ruslan Ermilov [Thu, 20 Oct 2005 10:45:19 +0000 (10:45 +0000)]
This commit was generated by cvs2svn to compensate for changes in r151497,
which included commits to RCS files with non-trunk default branches.

18 years agoVirgin import of FSF groff v1.19.2
Ruslan Ermilov [Thu, 20 Oct 2005 10:45:19 +0000 (10:45 +0000)]
Virgin import of FSF groff v1.19.2

18 years agoRemove unused variables.
Marius Strobl [Thu, 20 Oct 2005 10:39:09 +0000 (10:39 +0000)]
Remove unused variables.

18 years agoSome more minor cleanups of em(4) driver:
Gleb Smirnoff [Thu, 20 Oct 2005 09:55:49 +0000 (09:55 +0000)]
Some more minor cleanups of em(4) driver:
  - Destroy mutex in case of attach failure. [1]
  - Lock properly em_watchdog(). [1]
  - Lock properly em_sysctl_int_delay(). [1]
  - Remove unused global adapter linked list.
  - Remove unused dma_size field from struct em_dma_alloc.
  - Do not touch interface statistics, that must be edited
    only by upper layers. [1]

Submitted by: yongari [1]

18 years agoRevamp interrupt handling in em(4) driver:
Gleb Smirnoff [Thu, 20 Oct 2005 08:46:43 +0000 (08:46 +0000)]
Revamp interrupt handling in em(4) driver:

o Do not mask the RX overrun interrupt.

o Rewrite em_intr():
  - Axe EM_MAX_INTR.
  - Cycle acknowledging interrupts and processing
    packets until zero interrupt cause register is
    read.
  - If RX overrun comes in log this fact. [ NetBSD also
    resets adapter in this case, but my tests showed that
    this is not needed and only pessimizes behavior under
    heavy load. ]
  - Since almost all functions is rewritten, style the
    remaining lines.

This fixes em(4) interfaces wedging under high load.

In collaboration with: wpaul, cognet
Obtained from: NetBSD

18 years agoAdd el_get to the NAME section.
Stefan Farfeleder [Thu, 20 Oct 2005 08:26:03 +0000 (08:26 +0000)]
Add el_get to the NAME section.

Obtained from: OpenBSD (via NetBSD)

18 years agoFix a well duplicated fencepost error that stopped crashdumps being
Peter Wemm [Thu, 20 Oct 2005 05:41:38 +0000 (05:41 +0000)]
Fix a well duplicated fencepost error that stopped crashdumps being
readable on certain random memory configurations.  If the libkvm consumer
tried to read something that was in the very last pdpe, pde or pte slot,
it would bogusly fail.

This is broken in RELENG_6 too.

18 years agoSet the signal handlers before creating the thread. This is so it inherits
Nate Lawson [Thu, 20 Oct 2005 05:12:20 +0000 (05:12 +0000)]
Set the signal handlers before creating the thread.  This is so it inherits
the parent's signal mask.  Once daemon() forked, signals would be ignored
in the child thread.  While I'm here, check the return value of daemon().
This fixes termination in the daemon case (bug introduced in last commit).

Noticed by: Frederik Lindberg

18 years agowpa_supplicant(8) requires -D option for ndis(4) now.
Jung-uk Kim [Wed, 19 Oct 2005 22:26:47 +0000 (22:26 +0000)]
wpa_supplicant(8) requires -D option for ndis(4) now.

18 years agoMake devd WARNS=4 clean and bump WARNS accordingly. This will insure
Brooks Davis [Wed, 19 Oct 2005 21:21:22 +0000 (21:21 +0000)]
Make devd WARNS=4 clean and bump WARNS accordingly.  This will insure
that future variable shadowing bugs don't compile.

Reviewed by: imp
Compiled on: alpha i386 sparc64

18 years agoFix a bug in the kernel module runtime linker that made it impossible
John Polstra [Wed, 19 Oct 2005 20:40:30 +0000 (20:40 +0000)]
Fix a bug in the kernel module runtime linker that made it impossible
to unload the usb.ko module after boot if it was originally preloaded
from "/boot/loader.conf".  When processing preloaded modules, the
linker erroneously added self-dependencies the each module's reference
count.  That prevented usb.ko's reference count from ever going to 0,
so it could not be unloaded.

Sponsored by Isilon Systems.

Reviewed by: pjd, peter
MFC after: 1 week

18 years agoAs pointed out on current@, we don't want to declare a variable in a scope
Warner Losh [Wed, 19 Oct 2005 18:10:34 +0000 (18:10 +0000)]
As pointed out on current@, we don't want to declare a variable in a scope
that just uses the variable and throws it away.

This should fix the subsystem keyword wrt media-type.

MFC After: 2 days

18 years agosync with KAME (nuked unused code, use NULL to denote a NULL pointer)
SUZUKI Shinsuke [Wed, 19 Oct 2005 17:18:49 +0000 (17:18 +0000)]
sync with KAME (nuked unused code, use NULL to denote a NULL pointer)

Obtained from: KAME
Reviewed by: ume, gnn

18 years agosync with KAME (removed a unnecesary non-standard macro)
SUZUKI Shinsuke [Wed, 19 Oct 2005 16:53:24 +0000 (16:53 +0000)]
sync with KAME (removed a unnecesary non-standard macro)

Obtained from: KAME
Reviewd by: ume, gnn

18 years agosync with KAME regarding the following clarification in RFC3542:
SUZUKI Shinsuke [Wed, 19 Oct 2005 16:43:57 +0000 (16:43 +0000)]
sync with KAME regarding the following clarification in RFC3542:
 - disable IPv6 operation if DAD fails for some EUI-64 link-local addresses.
 - export get_hw_ifid() (and rename it) as a subroutine for this process.

Obtained from: KAME
Reviewd by: ume, gnn
MFC after: 2 week

18 years agoUse the new name H_SETSIZE instead of the old H_EVENT to set the history
Stefan Farfeleder [Wed, 19 Oct 2005 16:37:52 +0000 (16:37 +0000)]
Use the new name H_SETSIZE instead of the old H_EVENT to set the history
size.

PR: 86355
Approved by: gad

18 years agosync with KAME (don't respond to NI_QTYPE_IPV4ADDR)
SUZUKI Shinsuke [Wed, 19 Oct 2005 16:27:33 +0000 (16:27 +0000)]
sync with KAME (don't respond to NI_QTYPE_IPV4ADDR)

Obtained from: KAME
Reviewed by: ume, gnn

18 years agosupported an ndp command suboption to disable IPv6 in the given interface
SUZUKI Shinsuke [Wed, 19 Oct 2005 16:20:18 +0000 (16:20 +0000)]
supported an ndp command suboption to disable IPv6 in the given interface

Obtained from: KAME
Reviewd by: ume, gnn
MFC after: 2 week

18 years agosync with KAME in the following point:
SUZUKI Shinsuke [Wed, 19 Oct 2005 15:58:23 +0000 (15:58 +0000)]
sync with KAME in the following point:
- fixed a bug that "ndp ... proxy" command does not work

Obtained from: KAME
Reviewd by: ume, gnn
MFC after: 2 weeks

18 years agosync with KAME in the following points:
SUZUKI Shinsuke [Wed, 19 Oct 2005 15:52:00 +0000 (15:52 +0000)]
sync with KAME in the following points:
- removed compilation warnings
- suppress a redundant error message when a default-router-list is empty

Obtained from: KAME
Reviewd by: ume, gnn
MFC after: 2 weeks

18 years agoUse the new name H_SETSIZE instead of the old H_EVENT to set the history
Stefan Farfeleder [Wed, 19 Oct 2005 15:37:43 +0000 (15:37 +0000)]
Use the new name H_SETSIZE instead of the old H_EVENT to set the history
size.

PR: 86355

18 years agochanged syslog level to more appropriate ones
SUZUKI Shinsuke [Wed, 19 Oct 2005 15:19:06 +0000 (15:19 +0000)]
changed syslog level to more appropriate ones

Obtained from: KAME
Reviewd by: ume, gnn
MFC after: 2 weeks

18 years agosource link-layer address option should be marked to be checked later,
SUZUKI Shinsuke [Wed, 19 Oct 2005 15:14:28 +0000 (15:14 +0000)]
source link-layer address option should be marked to be checked later,
because rs_input() need this option.

Obtained from: KAME
Reviewed by: ume, gnn
MFC after: 2 weeks

18 years agoadded an ioctl option in kernel so that ndp/rtadvd can change some NDP-related kernel...
SUZUKI Shinsuke [Wed, 19 Oct 2005 15:05:42 +0000 (15:05 +0000)]
added an ioctl option in kernel so that ndp/rtadvd can change some NDP-related kernel variables based on their configurations (RFC2461 p.43 6.2.1 mandates this for IPv6 routers)

Obtained from: KAME
Reviewd by: ume, gnn
MFC after: 2 weeks

18 years agoUse svr4_si_{addr,code,errno,signo,trap} in preference to si_{...}.
Robert Watson [Wed, 19 Oct 2005 14:59:54 +0000 (14:59 +0000)]
Use svr4_si_{addr,code,errno,signo,trap} in preference to si_{...}.

Fix a debugging printf to printf after a variable is first assigned,
not before.

These are purely build fixes, and need inspection to make sure they
were what the original author of the previous changes intended.

18 years ago In the em_process_receive_interrupts() cycle check the IFF_DRV_RUNNING
Gleb Smirnoff [Wed, 19 Oct 2005 13:34:48 +0000 (13:34 +0000)]
  In the em_process_receive_interrupts() cycle check the IFF_DRV_RUNNING
flag. This fixes panic, when 'ifconfig em0 down' was called and it calls
em_stop() while the em_process_receive_interrupts() has temporarily
dropped the lock.

18 years agosync with KAME in the following points:
SUZUKI Shinsuke [Wed, 19 Oct 2005 10:09:19 +0000 (10:09 +0000)]
sync with KAME in the following points:
- fixed typos
- improved some comment descriptions
- use NULL, instead of 0, to denote a NULL pointer
- avoid embedding a magic number in the code
- use nd6log() instead of log() to record NDP-specific logs
- nuked an unnecessay white space

Obtained from: KAME
MFC after:  1 day

18 years agoConvert if (tp->t_state == TCPS_LISTEN) panic() into a KASSERT.
Robert Watson [Wed, 19 Oct 2005 09:37:52 +0000 (09:37 +0000)]
Convert if (tp->t_state == TCPS_LISTEN) panic() into a KASSERT.

MFC after: 2 weeks

18 years agoFix compiling problem by adding prefix name svr4 to si_xxx macro, the
David Xu [Wed, 19 Oct 2005 09:33:15 +0000 (09:33 +0000)]
Fix compiling problem by adding prefix name svr4 to si_xxx macro, the
si_xxx macro should not be used in compat headers, as these are standard
member names or only can be used in our native header file signal.h.

18 years agoFix typo
Poul-Henning Kamp [Wed, 19 Oct 2005 06:37:39 +0000 (06:37 +0000)]
Fix typo

18 years agoUse a pipe for reading devd events as another method of getting the AC line
Nate Lawson [Wed, 19 Oct 2005 04:48:44 +0000 (04:48 +0000)]
Use a pipe for reading devd events as another method of getting the AC line
status.  Add a thread that waits for events on the named pipe instead of
polling the line status via apm or the ACPI sysctl.  Additionally, use
robust error handling in case devd goes away or we temporarily can't set
a frequency (i.e., passive cooling has preempted our setting).

Later, this should be improved so that if adaptive control is not being used,
we will block while waiting for AC line events, saving a little CPU.

Submitted by: Frederik Lindberg <fli+freebsd-current/shapeshifter.se>

18 years agoCorrect the size used to allocate ndis_txarray. It should be
Bill Paul [Wed, 19 Oct 2005 03:14:11 +0000 (03:14 +0000)]
Correct the size used to allocate ndis_txarray. It should be
NDIS_TXPKTS now, not ndis_maxpkts.

18 years agoRaw IPv6 checksum must use the protocol number of the last header, instead of the...
SUZUKI Shinsuke [Wed, 19 Oct 2005 01:21:49 +0000 (01:21 +0000)]
Raw IPv6 checksum must use the protocol number of the last header, instead of the first next-header value.

Obtained from: KAME
MFC after: 1 day

18 years agoSet the name of the chipset found
Warner Losh [Wed, 19 Oct 2005 00:25:39 +0000 (00:25 +0000)]
Set the name of the chipset found

18 years agoMake sure we set bst and bsh in the softc.
Warner Losh [Wed, 19 Oct 2005 00:18:11 +0000 (00:18 +0000)]
Make sure we set bst and bsh in the softc.

This gets us probing, but not attaching to, ISA cards.  More work
needed since the ISA attach routine is return ENXIO right now :-)

18 years agoFix vchan speed for hardware with discrete (non-continuous)
Ariff Abdullah [Tue, 18 Oct 2005 21:33:51 +0000 (21:33 +0000)]
Fix vchan speed for hardware with discrete (non-continuous)
sampling rate:
- Improve vchan chn_setspeed() strategy. Try to avoid FEEDER_RATE
  on parent channel if the requested value is not supported
  by the hardware.
- Fix vchan default speed calculation. In any case, vchan should
  rely on parent bufsoft speed instead of bufhard since it is
  possible that the entire feeder chain might involve FEEDER_RATE.
  This is possible under extreme, rare condition if the above
  chn_setspeed() strategy failed.

Approved by: netchild (mentor)

18 years agoAdded missing comma. This fixes compilation if we need to enable
Ariff Abdullah [Tue, 18 Oct 2005 21:18:47 +0000 (21:18 +0000)]
Added missing comma. This fixes compilation if we need to enable
RATE_ASSERT debug macro.

Approved by: netchild (mentor)

18 years agoBruce Schneier is underrepresented in the fortune file.
Poul-Henning Kamp [Tue, 18 Oct 2005 20:37:44 +0000 (20:37 +0000)]
Bruce Schneier is underrepresented in the fortune file.

18 years agoUse correct cirteria for determining which directory entries we can
Poul-Henning Kamp [Tue, 18 Oct 2005 20:21:25 +0000 (20:21 +0000)]
Use correct cirteria for determining which directory entries we can
purge right away and which we merely can hide.

Beaten into my skull by: kris

18 years agoExport processor socket information. New environment variables are:
Jung-uk Kim [Tue, 18 Oct 2005 20:03:31 +0000 (20:03 +0000)]
Export processor socket information.  New environment variables are:

smbios.socket.enabled: number of enabled sockets
smbios.socket.populated: number of populated sockets

18 years agoAnother round of cleanups and fixes:
Bill Paul [Tue, 18 Oct 2005 19:52:15 +0000 (19:52 +0000)]
Another round of cleanups and fixes:

- Change ndis_return() from a DPC to a workitem so that it doesn't
  run at DISPATCH_LEVEL (with the dispatcher lock held).

- In if_ndis.c, submit packets to the stack via (*ifp->if_input)() in
  a workitem instead of doing it directly in ndis_rxeof(), because
  ndis_rxeof() runs in a DPC, and hence at DISPATCH_LEVEL. This
  implies that the 'dispatch level' mutex for the current CPU is
  being held, and we don't want to call if_input while holding
  any locks.

- Reimplement IoConnectInterrupt()/IoDisconnectInterrupt(). The original
  approach I used to track down the interrupt resource (by scanning
  the device tree starting at the nexus) is prone to problems when
  two devices share an interrupt. (E.g removing ndis1 might disable
  interrupts for ndis0.) The new approach is to multiplex all the
  NDIS interrupts through a common internal dispatcher (ntoskrnl_intr())
  and allow IoConnectInterrupt()/IoDisconnectInterrupt() to add or
  remove interrupts from the dispatch list.

- Implement KeAcquireInterruptSpinLock() and KeReleaseInterruptSpinLock().

- Change the DPC and workitem threads to use the KeXXXSpinLock
  API instead of mtx_lock_spin()/mtx_unlock_spin().

- Simplify the NdisXXXPacket routines by creating an actual
  packet pool structure and using the InterlockedSList routines
  to manage the packet queue.

- Only honor the value returned by OID_GEN_MAXIMUM_SEND_PACKETS
  for serialized drivers. For deserialized drivers, we now create
  a packet array of 64 entries. (The Microsoft DDK documentation
  says that for deserialized miniports, OID_GEN_MAXIMUM_SEND_PACKETS
  is ignored, and the driver for the Marvell 8335 chip, which is
  a deserialized miniport, returns 1 when queried.)

- Clean up timer handling in subr_ntoskrnl.

- Add the following conditional debugging code:
NTOSKRNL_DEBUG_TIMERS - add debugging and stats for timers
NDIS_DEBUG_PACKETS - add extra sanity checking for NdisXXXPacket API
NTOSKRNL_DEBUG_SPINLOCKS - add test for spinning too long

- In kern_ndis.c, always start the HAL first and shut it down last,
  since Windows spinlocks depend on it. Ntoskrnl should similarly be
  started second and shut down next to last.

18 years agoMove the initialization of the devmtx into the mutex_init() function
John Baldwin [Tue, 18 Oct 2005 18:27:44 +0000 (18:27 +0000)]
Move the initialization of the devmtx into the mutex_init() function
called during early init before cninit().

Tested on: i386, alpha, sparc64
Reviewed by: phk, imp
Reported by: Divacky Roman xdivac02 at stud dot fit dot vutbr dot cz
MFC after: 1 week

18 years agoMake __sem_timedwait() consistent with the sem_timedwait() prototype.
Stefan Farfeleder [Tue, 18 Oct 2005 17:24:03 +0000 (17:24 +0000)]
Make __sem_timedwait() consistent with the sem_timedwait() prototype.

18 years agoxterm-basic: add AX flag
Andrey A. Chernov [Tue, 18 Oct 2005 16:11:48 +0000 (16:11 +0000)]
xterm-basic: add AX flag
(the same in xterm-supplied termcap)

18 years agoImplement the full range of ISO9660 number conversion routines in iso.h.
Dag-Erling Smørgrav [Tue, 18 Oct 2005 13:35:08 +0000 (13:35 +0000)]
Implement the full range of ISO9660 number conversion routines in iso.h.

MFC after: 2 weeks

18 years agoRegenerate.
Stefan Farfeleder [Tue, 18 Oct 2005 11:49:14 +0000 (11:49 +0000)]
Regenerate.

18 years agoConst-qualify ksem_timedwait's parameter abstime as it's only passed in.
Stefan Farfeleder [Tue, 18 Oct 2005 11:46:24 +0000 (11:46 +0000)]
Const-qualify ksem_timedwait's parameter abstime as it's only passed in.

18 years agoIn this case, disable wrapping of fake opt_*.h targets in
Yaroslav Tykhiy [Tue, 18 Oct 2005 07:54:02 +0000 (07:54 +0000)]
In this case, disable wrapping of fake opt_*.h targets in
"if defined(KERNBUILDDIR)" for now since the new way of
building modules with the kernel can't handle dynamic SRCS
depending on build options yet.

18 years agoAdd first tests for semicolon handling. The first test fails. This is
Hartmut Brandt [Tue, 18 Oct 2005 07:28:09 +0000 (07:28 +0000)]
Add first tests for semicolon handling. The first test fails. This is
probably a bug in the parser.

18 years agoAdd more tests for escaped newline handling and fix a test that currently
Hartmut Brandt [Tue, 18 Oct 2005 07:20:14 +0000 (07:20 +0000)]
Add more tests for escaped newline handling and fix a test that currently
should fail because of a bug in the parser (test 2).

18 years agoDon't build dc-only pseudo phy devices with mii.ko anymore
Warner Losh [Tue, 18 Oct 2005 06:39:30 +0000 (06:39 +0000)]
Don't build dc-only pseudo phy devices with mii.ko anymore

18 years agoSpotted a call to a cdev-related function not wrapped
Yaroslav Tykhiy [Tue, 18 Oct 2005 06:38:14 +0000 (06:38 +0000)]
Spotted a call to a cdev-related function not wrapped
in "#ifdef KBD_INSTALL_CDEV".

Noticed by: glebius
Uncovered by: recent change to modules' Makefiles

18 years agoMove dc sources from pci and dev/mii into dev/dc.
Warner Losh [Tue, 18 Oct 2005 06:11:59 +0000 (06:11 +0000)]
Move dc sources from pci and dev/mii into dev/dc.

18 years agoMove dc sources from pci to dev/dc.
Warner Losh [Tue, 18 Oct 2005 06:11:08 +0000 (06:11 +0000)]
Move dc sources from pci to dev/dc.

18 years agoRemove dc-only pseudo phy devices after repo copy to dev/dc
Warner Losh [Tue, 18 Oct 2005 06:10:46 +0000 (06:10 +0000)]
Remove dc-only pseudo phy devices after repo copy to dev/dc

18 years agoRemove dc after repo copy.
Warner Losh [Tue, 18 Oct 2005 06:10:03 +0000 (06:10 +0000)]
Remove dc after repo copy.

18 years agoMake dc compile after repo-copy.
Warner Losh [Tue, 18 Oct 2005 06:09:42 +0000 (06:09 +0000)]
Make dc compile after repo-copy.

18 years agoHandle 32 bit executables better on AMD64 platforms.
Joseph Koshy [Tue, 18 Oct 2005 05:13:27 +0000 (05:13 +0000)]
Handle 32 bit executables better on AMD64 platforms.

18 years agoDocument pmcstat's inability to handle 32 bit executables on the
Joseph Koshy [Tue, 18 Oct 2005 04:40:11 +0000 (04:40 +0000)]
Document pmcstat's inability to handle 32 bit executables on the
AMD64 while I work on a fix.

Noticed by: ps

18 years ago- Use BUS_DMASYNC_PREWRITE in em_get_buf(), as the adapter is about to read
Olivier Houchard [Tue, 18 Oct 2005 00:42:10 +0000 (00:42 +0000)]
- Use BUS_DMASYNC_PREWRITE in em_get_buf(), as the adapter is about to read
the descriptors set.
- In em_process_receive_interrupts(), call bus_dmamap_sync() for the
descriptors set each time we modify one descriptor, instead of doing it only
at the function exit, to make sure the adapters know he can re-use the
descriptor.
This helps on arm with write-back data cache (and possibly on other arches
with bounce pages, I don't know) under heavy network load. Without this,
if we attempt to process more than num_rx_desc descriptors, the adapter
would just stop processing rx interrupts.

18 years agoRedo physical/logical CPU count.
Jung-uk Kim [Mon, 17 Oct 2005 23:23:20 +0000 (23:23 +0000)]
Redo physical/logical CPU count.

Suggested by: jhb

18 years agoAdd support for kernel modules with a single PT_LOAD section.
Peter Wemm [Mon, 17 Oct 2005 23:21:55 +0000 (23:21 +0000)]
Add support for kernel modules with a single PT_LOAD section.

While here, support up to four sections because it was trivial to do
and cheap. (One pointer per section).

For amd64 with "-fpic -shared" format .ko files, using a single PT_LOAD
section is important to avoid wasting about 1MB of KVM and physical ram
for the 'gap' between the two PT_LOAD sections.  amd64 normally uses
.o format kld files and isn't affected normally.  But -fpic -shared modules
are actually possible to produce and load...  (And with a bugfix to
binutils, we can build and use plain -shared .ko files without -fpic)

i386 only wastes 4K per .ko file, so that isn't such a big deal there.

18 years agoMicro optimization for context switch. Eliminate code for saving gs.base
David Xu [Mon, 17 Oct 2005 23:10:31 +0000 (23:10 +0000)]
Micro optimization for context switch. Eliminate code for saving gs.base
and fs.base. We always update pcb.pcb_gsbase and pcb.pcb_fsbase
when user wants to set them, in context switch routine, we only need to
write them into registers, we never have to read them out from registers
when thread is switched away. Since rdmsr is a serialization instruction,
micro benchmark shows it is worthy to do.

Reviewed by: peter, jhb

18 years agoHook ipfwpcap into the build tree.
Poul-Henning Kamp [Mon, 17 Oct 2005 20:47:44 +0000 (20:47 +0000)]
Hook ipfwpcap into the build tree.

18 years agoAdd a small tool which captures packets on a DIVERT socket and writes
Poul-Henning Kamp [Mon, 17 Oct 2005 20:27:15 +0000 (20:27 +0000)]
Add a small tool which captures packets on a DIVERT socket and writes
them as pcap (tcpdump) format data to a file or pipe.

Volunters wanted for: manpage
Contributed by: P Kern <pkern@cns.utoronto.ca>

18 years agoAllow the process name to be in square brackets ([]) in _find_processes().
John Baldwin [Mon, 17 Oct 2005 19:01:53 +0000 (19:01 +0000)]
Allow the process name to be in square brackets ([]) in _find_processes().

PR: conf/82430
Submitted by: Pavel Volkov pol at iib dot ru
MFC after: 1 week

18 years agoAnother bit of sx(4) removal.
John Baldwin [Mon, 17 Oct 2005 18:35:57 +0000 (18:35 +0000)]
Another bit of sx(4) removal.

18 years agoStrip the $a, $t and $d symbols if we're using DDB. There are useless and
Olivier Houchard [Mon, 17 Oct 2005 18:31:37 +0000 (18:31 +0000)]
Strip the $a, $t and $d symbols if we're using DDB. There are useless and
confusing in a backtrace.

18 years agoRemove a spurious newline. The TAILQ_INIT() is part of the 'faster tailq
John Baldwin [Mon, 17 Oct 2005 17:53:17 +0000 (17:53 +0000)]
Remove a spurious newline.  The TAILQ_INIT() is part of the 'faster tailq
deletion' example.

MFC after: 1 week

18 years agoDocument that changes to nsswitch.conf might be needed when
Bjoern A. Zeeb [Mon, 17 Oct 2005 17:01:54 +0000 (17:01 +0000)]
Document that changes to nsswitch.conf might be needed when
compiling with NO_NIS.
[ also see make.conf(5) and nsswitch.conf(5) ]

Reviewed by:    ru
Requested by:   dougb (and bin/87221)
MFC:            22 days

18 years agostyle(9) nitlet:
Warner Losh [Mon, 17 Oct 2005 16:51:06 +0000 (16:51 +0000)]
style(9) nitlet:

foo() {
}

to

foo()
{
}

18 years agoMake sure that files included using ".include <foo>" are really
Ruslan Ermilov [Mon, 17 Oct 2005 15:56:26 +0000 (15:56 +0000)]
Make sure that files included using ".include <foo>" are really
looked for in the system make file directory or in the specified
-m paths instead of always looking in the other -I and .PATH
specified paths.  (Commit log shamelessly stolen from NetBSD.)

Reviewed by: yar

18 years agoSplit displaying number of physical and logical cores.
Jung-uk Kim [Mon, 17 Oct 2005 15:51:28 +0000 (15:51 +0000)]
Split displaying number of physical and logical cores.

18 years agoObtain true uptime through clock_gettime(CLOCK_MONOTONIC, struct *timespec)
Andre Oppermann [Mon, 17 Oct 2005 15:37:22 +0000 (15:37 +0000)]
Obtain true uptime through clock_gettime(CLOCK_MONOTONIC, struct *timespec)
instead of subtracting 'bootime' from 'now'.

Sponsored by: TCP/IP Optimization Fundraise 2005

18 years agoRemove now redundant NO_SHARED.
Ruslan Ermilov [Mon, 17 Oct 2005 15:28:25 +0000 (15:28 +0000)]
Remove now redundant NO_SHARED.

Reviewed by: imp

18 years agoUse the clock count register as a timecounter, as it's more accurate.
Olivier Houchard [Mon, 17 Oct 2005 14:51:01 +0000 (14:51 +0000)]
Use the clock count register as a timecounter, as it's more accurate.

18 years agoAdd a valuable piece of Usenet advice about storing passwords. :-)
Philip Paeps [Mon, 17 Oct 2005 14:27:37 +0000 (14:27 +0000)]
Add a valuable piece of Usenet advice about storing passwords. :-)

Obtained from: jpd on comp.unix.bsd.freebsd.misc

18 years agofixed a kernel crash when IPv6 PIM-SM routing is enabled and a PIM register message...
SUZUKI Shinsuke [Mon, 17 Oct 2005 13:47:31 +0000 (13:47 +0000)]
fixed a kernel crash when IPv6 PIM-SM routing is enabled and a PIM register message is received

Obtained from: KAME
MFC After: 3 days

18 years agoIn __option_is_short(), avoid calling isprint() on key characters outside
Tim J. Robbins [Mon, 17 Oct 2005 12:20:05 +0000 (12:20 +0000)]
In __option_is_short(), avoid calling isprint() on key characters outside
the range allowed by that function, resulting in undefined behaviour.
Our undefined behaviour in multibyte locales happened to differ from
glibc's, resulting in errors parsing option strings.

Obtained from: Corinna Vinschen (Red Hat)

18 years agoMinor commentary correction
Warner Losh [Mon, 17 Oct 2005 08:20:44 +0000 (08:20 +0000)]
Minor commentary correction

18 years agoxterm-basic:
Andrey A. Chernov [Mon, 17 Oct 2005 06:09:56 +0000 (06:09 +0000)]
xterm-basic:
1) use standard vt100 as/ae instead of ^N/^O which not works in real xterm
2) Remove ^O from me, it not belongs there

18 years agoUnconditionally mount a CD9660 filesystem as read-only, instead of
Craig Rodrigues [Mon, 17 Oct 2005 03:29:53 +0000 (03:29 +0000)]
Unconditionally mount a CD9660 filesystem as read-only, instead of
returning EROFS if we forget to mount it as read-only.

18 years agoUse the actual sector size of the media instead of hard-coding it to 2048.
Craig Rodrigues [Mon, 17 Oct 2005 03:27:35 +0000 (03:27 +0000)]
Use the actual sector size of the media instead of hard-coding it to 2048.
This eliminates KASSERTs in GEOM if we accidentally mount an audio CD
as a cd9660 filesystem.

18 years agoUnconditionally mount a UDF filesystem as read-only, instead of
Craig Rodrigues [Mon, 17 Oct 2005 03:07:36 +0000 (03:07 +0000)]
Unconditionally mount a UDF filesystem as read-only, instead of
returning an EROFS if we forget to mount it as read-only.

18 years agoCheck against 8019 rather than 8029 for reading in the MAC addr since
Warner Losh [Mon, 17 Oct 2005 02:32:23 +0000 (02:32 +0000)]
Check against 8019 rather than 8029 for reading in the MAC addr since
the former is the ISA part, not the latter.

MFC After 6.0 is unfrozen (this bug doesn't exist in 6.0 because I didn't
MFC the rtl80x9 changes for ISA due to an error on my part)

18 years ago- Fix typo.
Florent Thoumie [Mon, 17 Oct 2005 00:04:35 +0000 (00:04 +0000)]
- Fix typo.

Approved by: ssouhlal
MFC after: 1 week

18 years agoUse __pid_t instead of pid_t for sigqueue as other functions.
David Xu [Sun, 16 Oct 2005 22:23:03 +0000 (22:23 +0000)]
Use __pid_t instead of pid_t for sigqueue as other functions.

18 years agoUpdate nwfs_lookup() to match the current cache_lookup() API.
Don Lewis [Sun, 16 Oct 2005 21:54:35 +0000 (21:54 +0000)]
Update nwfs_lookup() to match the current cache_lookup() API.
cache_lookup() has returned a ref'ed and locked vnode since
vfs_cache.c:1.96, dated Tue Mar 29 12:59:06 2005 UTC.  This change
is similar to the change made to smbfs_lookup() in smbfs_vnops.c:1.58.

Tested by: "Antony Mawer" ant AT mawer.org
MFC after: 2 weeks