]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoAlphabetical order is probably better.
jb [Thu, 3 Aug 2006 03:30:53 +0000 (03:30 +0000)]
Alphabetical order is probably better.

18 years agoAdd the library directory where DTrace library scripts live.
jb [Thu, 3 Aug 2006 03:28:03 +0000 (03:28 +0000)]
Add the library directory where DTrace library scripts live.

18 years agoFix re(4) breakge introduced in tree from rev 1.68.
yongari [Thu, 3 Aug 2006 00:15:19 +0000 (00:15 +0000)]
Fix re(4) breakge introduced in tree from rev 1.68.
This should fix incorrect configuration of station address on
big-endian architectures.

Reviewed by: wpaul
Tested on: sparc64

18 years agoMove destroying kqueue state from above pru_detach to below it in
rwatson [Wed, 2 Aug 2006 18:37:44 +0000 (18:37 +0000)]
Move destroying kqueue state from above pru_detach to below it in
sofree(), as a number of protocols expect to be able to call
soisdisconnected() during detach.  That may not be a good assumption,
but until I'm sure if it's a good assumption or not, allow it.

18 years agoDon't ignore errors from intr_event_add_handler().
jhb [Wed, 2 Aug 2006 17:50:31 +0000 (17:50 +0000)]
Don't ignore errors from intr_event_add_handler().

CID: 1516
Found by: Coverity Prevent (tm)

18 years ago- Use m_getcl(), m_get(), and m_gethdr() rather than the older macros for
jhb [Wed, 2 Aug 2006 17:41:58 +0000 (17:41 +0000)]
- Use m_getcl(), m_get(), and m_gethdr() rather than the older macros for
  alloc'ing mbufs so that there is less error handling required.
- Go ahead and account for the data space in the first mbuf before entering
  the loop to alloc more mbuf's.  This simplifies the loop logic and avoids
  confusing Coverity.

CID: 817
Reviewed by: sam
Tested by: pjd
Found by: Coverity Prevent (tm)

18 years agoRemove remnants of Alpha.
marcel [Wed, 2 Aug 2006 17:22:30 +0000 (17:22 +0000)]
Remove remnants of Alpha.

18 years agoRemove remnants of Alpha.
marcel [Wed, 2 Aug 2006 17:16:48 +0000 (17:16 +0000)]
Remove remnants of Alpha.

18 years agoDefine the additional page fault error codes that are implemented by amd64.
alc [Wed, 2 Aug 2006 16:24:23 +0000 (16:24 +0000)]
Define the additional page fault error codes that are implemented by amd64.

18 years agoChange two XXX's to two notes: the fact that SOCK_LOCK(so) ==
rwatson [Wed, 2 Aug 2006 16:23:52 +0000 (16:23 +0000)]
Change two XXX's to two notes: the fact that SOCK_LOCK(so) ==
SOCKBUF_LOCK(&so->so_rcv) is encoded, which is worth noting, but not a
bug.

18 years agoRemove call to soisdisconnected() in at_pcbdetach(): by the time the
rwatson [Wed, 2 Aug 2006 16:22:34 +0000 (16:22 +0000)]
Remove call to soisdisconnected() in at_pcbdetach(): by the time the
socket is being detached, there are no consumers left worth notifying
about the disconnect.

18 years agoMove soisdisconnected() in tcp_discardcb() to one of its calling contexts,
rwatson [Wed, 2 Aug 2006 16:18:05 +0000 (16:18 +0000)]
Move soisdisconnected() in tcp_discardcb() to one of its calling contexts,
tcp_twstart(), but not to the other, tcp_detach(), as the socket is
already being torn down and therefore there are no listeners.  This avoids
a panic if kqueue state is registered on the socket at close(), and
eliminates to XXX comments.  There is one case remaining in which
tcp_discardcb() reaches up to the socket layer as part of the TCP host
cache, which would be good to avoid.

Reported by: Goran Gajic <ggajic at afrodita dot rcub dot bg dot ac dot yu>

18 years agoFix some bugs in the previous revision (1.419). Don't perform extra
jhb [Wed, 2 Aug 2006 15:27:48 +0000 (15:27 +0000)]
Fix some bugs in the previous revision (1.419).  Don't perform extra
vfs_rel() on the mountpoint if the MAC checks fail in kern_statfs() and
kern_fstatfs().  Similarly, don't perform an extra vfs_rel() if we get
a doomed vnode in kern_fstatfs(), and handle the case of mp being NULL
(for some doomed vnodes) by conditionalizing the vfs_rel() in
kern_fstatfs() on mp != NULL.

CID: 1517
Found by: Coverity Prevent (tm) (kern_fstatfs())
Pointy hat to: jhb

18 years agoFix doxygen target.
netchild [Wed, 2 Aug 2006 15:26:15 +0000 (15:26 +0000)]
Fix doxygen target.

Submitted by: "Alexander Mogilny" <sg@astral.ntu-kpi.kiev.ua>
Pointy hat to: netchild

18 years agoFix build when .CURDIR != .OBJDIR.
netchild [Wed, 2 Aug 2006 15:24:59 +0000 (15:24 +0000)]
Fix build when .CURDIR != .OBJDIR.

Noticed by: "Alexander Mogilny" <sg@astral.ntu-kpi.kiev.ua>

18 years agoDo not leak memory while flushing rules.
oleg [Wed, 2 Aug 2006 14:58:51 +0000 (14:58 +0000)]
Do not leak memory while flushing rules.

Noticed by: yar
Approved by: glebius (mentor)
MFC after: 1 week

18 years agoRemove now unneeded ENOTCONN clause from SOCK_DGRAM side of uipc_send():
rwatson [Wed, 2 Aug 2006 14:30:58 +0000 (14:30 +0000)]
Remove now unneeded ENOTCONN clause from SOCK_DGRAM side of uipc_send():
we have to check it regardless of the target address, so don't check it
twice.

18 years agoAdd a new command-line option, -P, which allows the user to specify
bms [Wed, 2 Aug 2006 13:21:44 +0000 (13:21 +0000)]
Add a new command-line option, -P, which allows the user to specify
an arbitrary port to which syslog messages are to be sent.

PR: bin/54026
Submitted by: Oliver Fromme
MFC after: 10 days

18 years agoBlock a variety of signals which may afffect reboot(8), before killing
bms [Wed, 2 Aug 2006 13:05:38 +0000 (13:05 +0000)]
Block a variety of signals which may afffect reboot(8), before killing
init(8), to avoid losing a race to them and dying before being able
to call reboot(2).

PR: bin/64664
Submitted by: maxim
Obtained from: NetBSD
MFC after: 30 days

18 years agoRemove 'register'.
rwatson [Wed, 2 Aug 2006 13:01:58 +0000 (13:01 +0000)]
Remove 'register'.
Use ANSI C prototypes/function headers.
More deterministically line wrap comments.

18 years agoBlock SIGHUP before killing init(8), to avoid a race condition which may
bms [Wed, 2 Aug 2006 12:42:20 +0000 (12:42 +0000)]
Block SIGHUP before killing init(8), to avoid a race condition which may
kill the current process and hang the system when attempting reboot.

PR: bin/64664
Reviewed by: ssouhal, phk (historic)
MFC after: 30 days

18 years agoRegenerate.
davidxu [Wed, 2 Aug 2006 08:14:01 +0000 (08:14 +0000)]
Regenerate.

18 years agoDon't include sys/thr.h and umtx.h in sys/sysproto.h, it is unnecessary.
davidxu [Wed, 2 Aug 2006 08:09:24 +0000 (08:09 +0000)]
Don't include sys/thr.h and umtx.h in sys/sysproto.h, it is unnecessary.

18 years agodon't include sys/thr.h and sys/umtx.h, it is unnecessary.
davidxu [Wed, 2 Aug 2006 07:38:59 +0000 (07:38 +0000)]
don't include sys/thr.h and sys/umtx.h, it is unnecessary.

18 years agoINT_MAX is defined in file sys/limits.h, include the file now.
davidxu [Wed, 2 Aug 2006 07:34:51 +0000 (07:34 +0000)]
INT_MAX is defined in file sys/limits.h, include the file now.

18 years agoCleanup and add our standard SYNOPSIS section.
brueffer [Wed, 2 Aug 2006 07:02:19 +0000 (07:02 +0000)]
Cleanup and add our standard SYNOPSIS section.

18 years agoReplace hard-coded magic constants to system defined constants
yongari [Wed, 2 Aug 2006 05:28:52 +0000 (05:28 +0000)]
Replace hard-coded magic constants to system defined constants
(BUS_PROBE_DEFAULT, BUS_PROBE_GENERIC etc). These pseudo PHY
drivers were forgotten from the conversion due to the repo copy
to dc driver location.

18 years agoremove unneccessary null ptr check
sam [Wed, 2 Aug 2006 04:55:04 +0000 (04:55 +0000)]
remove unneccessary null ptr check

Coverity ID: 173918

18 years agoremove unnecessary null ptr check
sam [Wed, 2 Aug 2006 04:54:14 +0000 (04:54 +0000)]
remove unnecessary null ptr check

Coverity ID: 173911

18 years agoDocument the log_stp variable.
thompsa [Wed, 2 Aug 2006 04:11:25 +0000 (04:11 +0000)]
Document the log_stp variable.

18 years ago- Use the new bridgestp callback to once again flush our bridge routes when an
thompsa [Wed, 2 Aug 2006 03:54:28 +0000 (03:54 +0000)]
- Use the new bridgestp callback to once again flush our bridge routes when an
  interface is disabled.
- Log port changes to syslog, defaulting to off

18 years agoTell bridgestp that we are about to free the memory so it can cleanup.
thompsa [Wed, 2 Aug 2006 02:59:24 +0000 (02:59 +0000)]
Tell bridgestp that we are about to free the memory so it can cleanup.

18 years agoFix style in the last commit, the variable declaration goes at the top of the
thompsa [Wed, 2 Aug 2006 02:51:42 +0000 (02:51 +0000)]
Fix style in the last commit, the variable declaration goes at the top of the
function.

18 years agoAdd a callback so we can notify the parent bridge that a port state change has
thompsa [Wed, 2 Aug 2006 02:47:27 +0000 (02:47 +0000)]
Add a callback so we can notify the parent bridge that a port state change has
occured, we need to do this from a taskqueue to avoid a LOR with the if_bridge
mutex.

18 years agoFix incorrect busy check for PHY write operation.
yongari [Wed, 2 Aug 2006 02:36:59 +0000 (02:36 +0000)]
Fix incorrect busy check for PHY write operation.
While I'm here remove unnecessary return statement.

18 years agoBe sure to disable the port when removing it from STP.
thompsa [Wed, 2 Aug 2006 01:36:40 +0000 (01:36 +0000)]
Be sure to disable the port when removing it from STP.

18 years agoMove updated of 'numopensockets' from bottom of sodealloc() to the top,
rwatson [Wed, 2 Aug 2006 00:45:27 +0000 (00:45 +0000)]
Move updated of 'numopensockets' from bottom of sodealloc() to the top,
eliminating a second set of identical mutex operations at the bottom.
This allows brief exceeding of the max sockets limit, but only by
sockets in the last stages of being torn down.

18 years agoDon't use f-word in comments. We are gentlemans.
pjd [Tue, 1 Aug 2006 23:17:33 +0000 (23:17 +0000)]
Don't use f-word in comments. We are gentlemans.

Pointed out by: Maciej Sobczak

18 years agoFix two nits in the ps header that offset each other making them largely
jhb [Tue, 1 Aug 2006 22:30:55 +0000 (22:30 +0000)]
Fix two nits in the ps header that offset each other making them largely
unnoticable.

18 years agoAdd device to access and modify Open Firmware NVRAM settings in
sobomax [Tue, 1 Aug 2006 22:19:01 +0000 (22:19 +0000)]
Add device to access and modify Open Firmware NVRAM settings in
PowerPC-based Apple's machines and small utility to do it from
userland modelled after the similar utility in Darwin/OSX.

Only tested on 1.25GHz G4 Mac Mini.

MFC after: 1 month

18 years agoComplete the transition from pmap_page_protect() to pmap_remove_write().
alc [Tue, 1 Aug 2006 19:06:06 +0000 (19:06 +0000)]
Complete the transition from pmap_page_protect() to pmap_remove_write().
Originally, I had adopted sparc64's name, pmap_clear_write(), for the
function that is now pmap_remove_write().  However, this function is more
like pmap_remove_all() than like pmap_clear_modify() or
pmap_clear_reference(), hence, the name change.

The higher-level rationale behind this change is described in
src/sys/amd64/amd64/pmap.c revision 1.567.  The short version is that I'm
trying to clean up and fix our support for execute access.

Reviewed by: marcel@ (ia64)

18 years agoIn vlan_input(), if the network interface does not perform h/w based
qingli [Tue, 1 Aug 2006 17:28:10 +0000 (17:28 +0000)]
In vlan_input(), if the network interface does not perform h/w based
vlan tag processing, the code will use bcopy() to remove the vlan
tag field but the code copies 2 bytes too many, which essentially
overwrites the protocol type field.

Also, a tag value of -1 is generated for unrecognized interface type,
which would cause an invalid memory access in the vlans[] array.

In addition, removed a line of dead code and its associated comments.

Reviewed by: sam

18 years agoAnother small update to the re(4) driver:
wpaul [Tue, 1 Aug 2006 17:18:25 +0000 (17:18 +0000)]
Another small update to the re(4) driver:

- Change the workaround for the autopad/checksum offload bug so that
  instead of lying about the map size, we actually create a properly
  padded mbuf and map it as usual. The other trick works, but is ugly.
  This approach also gives us a chance to zero the pad space to avoid
  possibly leaking data.

- With the PCIe devices, it looks issuing a TX command while there's
  already a transmission in progress doesn't have any effect. In other
  words, if you send two packets in rapid succession, the second one may
  end up sitting in the TX DMA ring until another transmit command is
  issued later in the future. Basically, if re_txeof() sees that there
  are still descriptors outstanding, it needs to manually resume the
  TX DMA channel by issuing another TX command to make sure all
  transmissions are flushed out. (The PCI devices seem to keep the
  TX channel moving until all descriptors have been consumed. I'm not
  sure why the PCIe devices behave differently.)

  (You can see this issue if you do the following test: plug an re(4)
  interface into another host via crossover cable, and from the other
  host do 'ping -c 2 <host with re(4) NIC>' to prime the ARP cache,
  then do 'ping -c 1 -s 1473 <host with re(4) NIC>'. You're supposed
  to see two packets sent in response, but you may only see one. If
  you do 'ping -c 1 -s 1473 <host with re(4) NIC>' again, you'll
  see two packets, but one will be the missing fragment from the last
  ping, followed by one of the fragments from this ping.)

- Add the PCI ID for the US Robotics 997902 NIC, which is based on
  the RTL8169S.

- Add a tsleep() of 1 second in re_detach() after the interrupt handler
  is disconnected. This should allow any tasks queued up by the ISR
  to drain. Now, I know you're supposed to use taskqueue_drain() for
  this, but something about the way taskqueue_drain() works with
  taskqueue_fast queues doesn't seem quite right, and I refuse to be
  tricked into fixing it.

18 years agoMake system call modules a bit more robust:
jhb [Tue, 1 Aug 2006 16:32:20 +0000 (16:32 +0000)]
Make system call modules a bit more robust:
- If we fail to register the system call during MOD_LOAD, then note that
  so that we don't try to deregister it or invoke the chained event handler
  during the subsequent MOD_UNLOAD event.  Doing the deregister when the
  register failed could result in trashing system call entries.
- Add a SI_SUB_SYSCALLS just before starting up init and use that to
  register syscall modules instead of SI_SUB_DRIVERS.  Registering system
  calls as late as possible increases the chances that any other module
  event handlers or SYSINITs in a module are executed to initialize the
  data in a kld before a syscall dependent on that data is able to be
  invoked.

MFC after: 3 days

18 years ago- Add a new function nfsrv_destroycache() to tear down the server request
jhb [Tue, 1 Aug 2006 16:27:14 +0000 (16:27 +0000)]
- Add a new function nfsrv_destroycache() to tear down the server request
  cache when unloading the nfsserver module.  This fixes a memory leak and
  a stale pointer.
- Use callout_drain() rather than callout_stop() when unloading the
  nfsserver module.

MFC after: 3 days

18 years agoUse TAILQ_FOREACH_SAFE() in a couple of places.
jhb [Tue, 1 Aug 2006 15:32:25 +0000 (15:32 +0000)]
Use TAILQ_FOREACH_SAFE() in a couple of places.

18 years agoDon't lock each of the processes while looking for a pid. The allproc and
jhb [Tue, 1 Aug 2006 15:30:56 +0000 (15:30 +0000)]
Don't lock each of the processes while looking for a pid.  The allproc and
proctree locks that we already hold provide sufficient protection.

18 years agoSome cosmetic tweaks:
jhb [Tue, 1 Aug 2006 15:29:46 +0000 (15:29 +0000)]
Some cosmetic tweaks:
- Right justify 'pid' label.
- Move the uid column to the right 2 columns so that the 3 process id
  columns (pid, ppid, pgrp) are grouped together.
- Expand the uid column to 5 chars.
- Don't indent the tid for multithreaded processes.

Requested by: bde (1, 2, 4)

18 years agoo Reserve space for 3.x releases similar to 2.1.
maxim [Tue, 1 Aug 2006 15:26:09 +0000 (15:26 +0000)]
o Reserve space for 3.x releases similar to 2.1.
o 3.0.1 is a security/critical release.

Submitted by: rpaulo@netbsd

18 years ago- add read only sysctl to indicate if write-combining was enabled
gallatin [Tue, 1 Aug 2006 14:02:54 +0000 (14:02 +0000)]
- add read only sysctl to indicate if write-combining was enabled
- enable mxge_dummy_rdma() right after reset, and make sure to disable
  when detaching the driver.

18 years agoReimplement socket buffer tear-down in sofree(): as the socket is no
rwatson [Tue, 1 Aug 2006 10:30:26 +0000 (10:30 +0000)]
Reimplement socket buffer tear-down in sofree(): as the socket is no
longer referenced by other threads (hence our freeing it), we don't need
to set the can't send and can't receive flags, wake up the consumers,
perform two levels of locking, etc.  Implement a fast-path teardown,
sbdestroy(), which flushes and releases each socket buffer.  A manual
dom_dispose of the receive buffer is still required explicitly to GC
any in-flight file descriptors, etc, before flushing the buffer.

This results in a 9% UP performance improvement and 16% SMP performance
improvement on a tight loop of socket();close(); in micro-benchmarking,
but will likely also affect CPU-bound macro-benchmark performance.

18 years agoIf skip_file_dev and skip_file_ino haven't been set (are still == 0),
kientzle [Tue, 1 Aug 2006 05:31:29 +0000 (05:31 +0000)]
If skip_file_dev and skip_file_ino haven't been set (are still == 0),
then don't use them for testing for a recursive add.

Thanks to: Spencer Minear
MFC after: 7 days

18 years agoCorrect spelling of 3DNow!.
obrien [Tue, 1 Aug 2006 01:23:39 +0000 (01:23 +0000)]
Correct spelling of 3DNow!.

18 years agoClose a race that occurs when using sendto() to connect and send on a
rwatson [Mon, 31 Jul 2006 23:00:05 +0000 (23:00 +0000)]
Close a race that occurs when using sendto() to connect and send on a
UNIX domain socket at the same time as the remote host is closing the
new connections as quickly as they open.  Since the connect() and
send() paths are non-atomic with respect to another, it is possible
for the second thread's close() call to disconnect the two sockets
as connect() returns, leading to the consumer (which plans to send())
with a NULL kernel pointer to its proposed peer.  As a result, after
acquiring the UNIX domain socket subsystem lock, we need to revalidate
the connection pointers even though connect() has technically succeed,
and reurn an error to say that there's no connection on which to
perform the send.

We might want to rethink the specific errno number, perhaps ECONNRESET
would be better.

PR: 100940
Reported by: Young Hyun <youngh at caida dot org>
MFC after: 2 weeks
MFC note: Some adaptation will be required

18 years agoAdd some statistics that are needed to support RFC4188 as part of the SoC2006
thompsa [Mon, 31 Jul 2006 20:24:46 +0000 (20:24 +0000)]
Add some statistics that are needed to support RFC4188 as part of the SoC2006
work on a bridge monitoring module for BSNMP.

Submitted by: shteryana (SoC 2006)

18 years agoinetd and telnetd are not included in the standard release
yar [Mon, 31 Jul 2006 19:15:10 +0000 (19:15 +0000)]
inetd and telnetd are not included in the standard release
crunched floppies, but they can be included as options in
src/release/picobsd (omitted by default though.)  Therefore
preserve the RELEASE_CRUNCH knob in their Makefiles, but
tell its real purpose in a comment.

18 years agoNote hardware supported by powerpc, based on GENERIC.
brueffer [Mon, 31 Jul 2006 17:04:28 +0000 (17:04 +0000)]
Note hardware supported by powerpc, based on GENERIC.

18 years agoMark two more items as merged.
brueffer [Mon, 31 Jul 2006 16:49:25 +0000 (16:49 +0000)]
Mark two more items as merged.

18 years agoRather than print out a nice error message giving details sufficent to fix
obrien [Mon, 31 Jul 2006 15:44:13 +0000 (15:44 +0000)]
Rather than print out a nice error message giving details sufficent to fix
a 'ufs_dirbad' and then panicing (making it very hard to see the details),
put them in the panic message itself.

18 years agoTrim an obsolete comment. ktrgenio() stopped doing crazy gymnastics when
jhb [Mon, 31 Jul 2006 15:31:43 +0000 (15:31 +0000)]
Trim an obsolete comment.  ktrgenio() stopped doing crazy gymnastics when
ktrace was redone to be mostly synchronous again.

18 years agoIn 'od -c' mode, deal with printable but zero-width combining
jkoshy [Mon, 31 Jul 2006 14:17:04 +0000 (14:17 +0000)]
In 'od -c' mode, deal with printable but zero-width combining
characters correctly.  These characters are displayed "combined"
with a space character.

PR: misc/100215
Submitted by: "J.R. Oldroyd" <<fbsd AT opal.com>>
Reviewed by: "J.R. Oldroyd" <<fbsd AT opal.com>> (revised patch)
MFC after: 3 days

18 years agoRegen.
yar [Mon, 31 Jul 2006 14:13:25 +0000 (14:13 +0000)]
Regen.

18 years agoTell the bitter truth that WITHOUT_GNU and WITHOUT_PAM
yar [Mon, 31 Jul 2006 13:53:27 +0000 (13:53 +0000)]
Tell the bitter truth that WITHOUT_GNU and WITHOUT_PAM
don't have any effect now.

18 years agoStop enforcing dependencies between MK_* options at Makefile level.
yar [Mon, 31 Jul 2006 13:29:36 +0000 (13:29 +0000)]
Stop enforcing dependencies between MK_* options at Makefile level.
All the dependencies are satisfied now in <bsd.own.mk>.

18 years agotelnetd(8) doesn't really go to the crunched floppies,
yar [Mon, 31 Jul 2006 13:20:44 +0000 (13:20 +0000)]
telnetd(8) doesn't really go to the crunched floppies,
so its Makefile needn't test for RELEASE_CRUNCH.

Suggested by: ru

18 years agoRegen from src/tools/build/options.
yar [Mon, 31 Jul 2006 13:15:04 +0000 (13:15 +0000)]
Regen from src/tools/build/options.

18 years agoDocument some more src.conf(5) options:
yar [Mon, 31 Jul 2006 13:08:47 +0000 (13:08 +0000)]
Document some more src.conf(5) options:
WITHOUT_GNU, WITHOUT_NS_CACHING, WITHOUT_PAM.

Noticed by: src/tools/build/options/makeman

18 years agoActually skip over undocumented options with "continue"
yar [Mon, 31 Jul 2006 12:41:13 +0000 (12:41 +0000)]
Actually skip over undocumented options with "continue"
to avoid artifacts in the manpage generated.  Previously
an orphaned paragraph on dependencies of such an option
would appear.

18 years agoDo not forget to increment the input line counter
yar [Mon, 31 Jul 2006 11:32:12 +0000 (11:32 +0000)]
Do not forget to increment the input line counter
when reading a word spanning multiple lines.

PR: bin/101094
MFC after: 5 days

18 years agoGranting the amount of misunderstanding the last change received,
yar [Mon, 31 Jul 2006 09:25:20 +0000 (09:25 +0000)]
Granting the amount of misunderstanding the last change received,
extend it with an example to clarify the point.

18 years agoRemove two unused variables.
kientzle [Mon, 31 Jul 2006 04:57:46 +0000 (04:57 +0000)]
Remove two unused variables.

Thanks to: Stefan Farfeleder

18 years agoAdd the bones for PowerPC release documentation. This has been
marcel [Mon, 31 Jul 2006 01:32:30 +0000 (01:32 +0000)]
Add the bones for PowerPC release documentation. This has been
derived from ia64.

18 years agoAdd proc-ia64.sgml to SRCS (uncomment).
marcel [Mon, 31 Jul 2006 01:27:16 +0000 (01:27 +0000)]
Add proc-ia64.sgml to SRCS (uncomment).

18 years agoFix the following bugs in re(4)
wpaul [Sun, 30 Jul 2006 23:25:21 +0000 (23:25 +0000)]
Fix the following bugs in re(4)

- Correct the PCI ID for the 8169SC/8110SC in the device list (I added
  the macro for it to if_rlreg.h before, but forgot to use it.)

- Remove the extra interrupt spinlock I added previously. After giving it
  some more thought, it's not really needed.

- Work around a hardware bug in some versions of the 8169. When sending
  very small IP datagrams with checksum offload enabled, a conflict can
  occur between the TX autopadding feature and the hardware checksumming
  that can corrupt the outbound packet. This is the reason that checksum
  offload sometimes breaks NFS: if you're using NFS over UDP, and you're
  very unlucky, you might find yourself doing a fragmented NFS write where
  the last fragment is smaller than the minimum ethernet frame size (60
  bytes). (It's rare, but if you keep NFS running long enough it'll
  happen.) If checksum offload is enabled, the chip will have to both
  autopad the fragment and calculate its checksum header. This confuses
  some revs of the 8169, causing the packet that appears on the wire
  to be corrupted. (The IP addresses and the checksum field are mangled.)
  This will cause the NFS write to fail. Unfortunately, when NFS retries,
  it sends the same write request over and over again, and it keeps
  failing, so NFS stays wedged.

  (A simple way to provoke the failure is to connect the failing system
  to a network with a known good machine and do "ping -s 1473 <badhost>"
  from the good system. The ping will fail.)

  Someone had previously worked around this using the heavy-handed
  approahch of just disabling checksum offload. The correct fix is to
  manually pad short frames where the TCP/IP stack has requested
  checksum offloading. This allows us to have checksum offload turned
  on by default but still let NFS work right.

- Not a bug, but change the ID strings for devices with hardware rev
  0x30000000 and 0x38000000 to both be 8168B/8111B. According to RealTek,
  they're both the same device, but 0x30000000 is an earlier silicon spin.

18 years agoUse NO_SHARED=YES to force a static link.
marcel [Sun, 30 Jul 2006 20:51:41 +0000 (20:51 +0000)]
Use NO_SHARED=YES to force a static link.

Pointed out by: ru@

18 years agoPass BN_CTX to internal functions instead of allocating it internally.
simon [Sun, 30 Jul 2006 19:29:26 +0000 (19:29 +0000)]
Pass BN_CTX to internal functions instead of allocating it internally.
This allows msqrt() to only call BN_CTX_new() once intead of many times.

Suggested and reviewed by: stefanf

18 years agoConditionally include sys/mkdev.h on platforms (such as Solaris) that need it.
kientzle [Sun, 30 Jul 2006 18:34:40 +0000 (18:34 +0000)]
Conditionally include sys/mkdev.h on platforms (such as Solaris) that need it.
While I'm here, correctly alphabetize 't' after 's'.

Thanks to: VMiklos

18 years agoConditionally include sys/mkdev.h on platforms (such as Solaris) that need it.
kientzle [Sun, 30 Jul 2006 18:33:20 +0000 (18:33 +0000)]
Conditionally include sys/mkdev.h on platforms (such as Solaris) that need it.

Thanks to: VMiklos

18 years agoResolve conflicts after import of OpenSSL 0.9.8b.
simon [Sun, 30 Jul 2006 14:17:54 +0000 (14:17 +0000)]
Resolve conflicts after import of OpenSSL 0.9.8b.

This was missed the first time around since eng_padlock.c was not part
of OpenSSL 0.9.7e and therefor did not have the v0_9_7e CVS tag used
during original resolve of conflicts.

Noticed by: Antoine Brodin <antoine.brodin@laposte.net>

18 years agoThis script should probably have an enabling variable since it can produce
mckay [Sun, 30 Jul 2006 12:54:37 +0000 (12:54 +0000)]
This script should probably have an enabling variable since it can produce
surprising results.  For now, at least make it safe to boot the default
kernel when /boot/kernel is already a symlink.

18 years agoo Kill unneeded assignment.
maxim [Sun, 30 Jul 2006 07:41:28 +0000 (07:41 +0000)]
o Kill unneeded assignment.

Obtained from: OpenBSD

18 years agoUse 'skip' when ignoring data in tar archives. This dramatically
kientzle [Sun, 30 Jul 2006 00:29:01 +0000 (00:29 +0000)]
Use 'skip' when ignoring data in tar archives.  This dramatically
increases performance when extracting a single entry from a large
uncompressed archive, especially on slow devices such as USB hard
drives.

Requires a number of changes:
   * New archive_read_open2() supports a 'skip' client function
   * Old archive_read_open() is implemented as a wrapper now, to
     continue supporting the old API/ABI.
   * _read_open_fd and _read_open_file sprout new 'skip' functions.
   * compression layer gets a new 'skip' operation.
   * compression_none passes skip requests through to client.
   * compression_{gzip,bzip2,compress} simply ignore skip requests.

Thanks to: Benjamin Lutz, who designed and implemented the whole thing.
   I'm just committing it.  ;-)

TODO: Need to update the documentation a little bit.

18 years agoDon't mention 'pax' in the context of POSIX-1988, since
kientzle [Sat, 29 Jul 2006 23:51:10 +0000 (23:51 +0000)]
Don't mention 'pax' in the context of POSIX-1988, since
pax wasn't introduced until the 1993 (?) revision.

(I need to double-check when pax was introduced and
clarify some of the history here.  In particular,
I should explain that the 'pax' standard now owns the
'ustar' format spec.)

18 years agoRemove mention of 'tp' format, since that support has been
kientzle [Sat, 29 Jul 2006 23:49:25 +0000 (23:49 +0000)]
Remove mention of 'tp' format, since that support has been
removed.  (It was introduced experimentally and I have simply
never had time to finish it.)

18 years agoSync FREEBSD-Xlist with what was actually excluded from OpenSSL 0.9.8b
simon [Sat, 29 Jul 2006 22:40:45 +0000 (22:40 +0000)]
Sync FREEBSD-Xlist with what was actually excluded from OpenSSL 0.9.8b
import.

18 years agoAdd some rough notes on how to import a new OpenSSL version into the
simon [Sat, 29 Jul 2006 22:01:26 +0000 (22:01 +0000)]
Add some rough notes on how to import a new OpenSSL version into the
FreeBSD base system.  Parts are inspired by the OpenSSH upgrade notes.

18 years agoDocument many of the sysctls in acpi. Bump the date on acpi_thermal.4
njl [Sat, 29 Jul 2006 21:47:24 +0000 (21:47 +0000)]
Document many of the sysctls in acpi.  Bump the date on acpi_thermal.4
because I forgot to on the last change.

MFC after: 1 day

18 years agoAdd a new sysctl, hw.acpi.handle_reboot. If set, acpi will attempt to
njl [Sat, 29 Jul 2006 21:46:16 +0000 (21:46 +0000)]
Add a new sysctl, hw.acpi.handle_reboot.  If set, acpi will attempt to
perform the reboot action via the reset register instead of our legacy
method.  Default is 0 (use legacy).  This is needed because some systems
hang on reboot even though they claim to support the reset register.

MFC after: 2 days

18 years agoAdd /usr/lib/engines for OpenSSL engines.
simon [Sat, 29 Jul 2006 19:47:09 +0000 (19:47 +0000)]
Add /usr/lib/engines for OpenSSL engines.

18 years agoBump __FreeBSD_version for OpenSSL 0.9.8b import.
simon [Sat, 29 Jul 2006 19:44:07 +0000 (19:44 +0000)]
Bump __FreeBSD_version for OpenSSL 0.9.8b import.

18 years agoLink kldxref(8) static on PowerPC to work around a SIGSEGV that
marcel [Sat, 29 Jul 2006 19:43:26 +0000 (19:43 +0000)]
Link kldxref(8) static on PowerPC to work around a SIGSEGV that
cannot easily be analyzed due to there being no debugger yet.
The SIGSEGV only happens when kldxref is linked shared.
Since kldxref(8) is needed for a release build, having it not
dump core is important.

18 years agoUpgrade to OpenSSL 0.9.8b.
simon [Sat, 29 Jul 2006 19:41:41 +0000 (19:41 +0000)]
Upgrade to OpenSSL 0.9.8b.

18 years agoChange maketempfile() to return a FILE* so as to eliminate the fopen()
marcel [Sat, 29 Jul 2006 19:39:03 +0000 (19:39 +0000)]
Change maketempfile() to return a FILE* so as to eliminate the fopen()
that immediately follows the only call to it. maketempfile() uses
mkstemp(), so the temporary file has already been opened and using
fopen() again just opens the file twice. This also fixes the invalid
mode used on the fopen().
While here, assign NULL to fxref after fclose() because we test for
fxref being !NULL to determine if we have the (temporary) hints file
open.

18 years agoResolve conflicts after import of OpenSSL 0.9.8b.
simon [Sat, 29 Jul 2006 19:14:51 +0000 (19:14 +0000)]
Resolve conflicts after import of OpenSSL 0.9.8b.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r160814,
simon [Sat, 29 Jul 2006 19:10:21 +0000 (19:10 +0000)]
This commit was generated by cvs2svn to compensate for changes in r160814,
which included commits to RCS files with non-trunk default branches.

18 years agoVendor import of OpenSSL 0.9.8b
simon [Sat, 29 Jul 2006 19:10:21 +0000 (19:10 +0000)]
Vendor import of OpenSSL 0.9.8b

18 years agoRemove sio(4) and related options from MI files to amd64, i386
marcel [Sat, 29 Jul 2006 18:38:54 +0000 (18:38 +0000)]
Remove sio(4) and related options from MI files to amd64, i386
and pc98 MD files. Remove nodevice and nooption lines specific
to sio(4) from ia64, powerpc and sparc64 NOTES. There were no
such lines for arm yet.
sio(4) is usable on less than half the platforms, not counting
a future mips platform. Its presence in MI files is therefore
increasingly becoming a burden.

18 years agoTell more of the sh(1) history.
yar [Sat, 29 Jul 2006 09:56:29 +0000 (09:56 +0000)]
Tell more of the sh(1) history.
Acknowledge Kenneth Almquist's contribution in AUTHORS.

MFC after: 5 days

18 years agoExtend the MK_INET6 section by rescue/ping6 and the manpages.
yar [Sat, 29 Jul 2006 06:29:07 +0000 (06:29 +0000)]
Extend the MK_INET6 section by rescue/ping6 and the manpages.

18 years agoDo not put BN_CTX structures on the stack, but instead allocate them
simon [Fri, 28 Jul 2006 23:00:16 +0000 (23:00 +0000)]
Do not put BN_CTX structures on the stack, but instead allocate them
runtime using BN_CTX_new().  This is done since in OpenSSL 0.9.7e we
can only allocate BN_CTX on the stack by including an internal OpenSSL
header file, and in OpenSSL 0.9.8 BN_CTX is entirely opaque, so having
it on the stack is not possible at all.

This is done as preparation for OpenSSL 0.9.8b import.

Tested on: amd64 i386 ia64
Tested with: src/tools/regression/lib/libmp