]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years agoRemove extra parenthesis used in macros. These macros are not used
yongari [Mon, 18 May 2009 07:13:42 +0000 (07:13 +0000)]
Remove extra parenthesis used in macros. These macros are not used
in driver, though.

15 years agoConsistently use tab characters instead of spaces.
yongari [Mon, 18 May 2009 07:10:48 +0000 (07:10 +0000)]
Consistently use tab characters instead of spaces.

15 years agostyle(9)
yongari [Mon, 18 May 2009 07:04:03 +0000 (07:04 +0000)]
style(9)

15 years agos/u_int8_t/uint8_t/g
yongari [Mon, 18 May 2009 06:32:38 +0000 (06:32 +0000)]
s/u_int8_t/uint8_t/g
s/u_int16_t/uint16_t/g
s/u_int32_t/uint32_t/g

15 years agostyle(9) - space after keyword.
yongari [Mon, 18 May 2009 06:19:17 +0000 (06:19 +0000)]
style(9) - space after keyword.

15 years agoRemove return statement at the end of functions that return void.
yongari [Mon, 18 May 2009 06:13:56 +0000 (06:13 +0000)]
Remove return statement at the end of functions that return void.

15 years agoRemove trailling whitespaces.
yongari [Mon, 18 May 2009 06:09:10 +0000 (06:09 +0000)]
Remove trailling whitespaces.

15 years agoRemove register keyword.
yongari [Mon, 18 May 2009 06:05:50 +0000 (06:05 +0000)]
Remove register keyword.

15 years agoUse ANSI C declarations for all functions.
yongari [Mon, 18 May 2009 06:02:54 +0000 (06:02 +0000)]
Use ANSI C declarations for all functions.

15 years agoThe merge in r189699 reverted part of the work done in a previous commit
adrian [Mon, 18 May 2009 04:56:37 +0000 (04:56 +0000)]
The merge in r189699 reverted part of the work done in a previous commit
(r188036.)

Re-revert that change so the Xen networking functions again.

15 years agoDisable some un-needed console debugging.
adrian [Mon, 18 May 2009 04:50:31 +0000 (04:50 +0000)]
Disable some un-needed console debugging.

15 years agoImplement MSG_CMSG_CLOEXEC flag for linux_recvmsg().
dchagin [Mon, 18 May 2009 04:07:46 +0000 (04:07 +0000)]
Implement MSG_CMSG_CLOEXEC flag for linux_recvmsg().

Approved by: kib (mentor)
MFC after: 1 month

15 years agoThis patch resolves the following issues:
qingli [Mon, 18 May 2009 02:25:45 +0000 (02:25 +0000)]
This patch resolves the following issues:

-- A routing socket message is not generated when an IPv6 address is
   either inserted or deleted from an interface. The missing routing
   message problem was discovered by Randall Stewart and Michael Tuxen
   during SCTP testing.

-- Previously when an IPv6 address is configured on an interface, if the
   prefix length is /128, then a host route is instaleld in the kernel
   for this address. But this host route is not deleted when that IPv6
   address is removed from the interface.

-- Routes to the link-local all-nodes multicast address and the
   interface-local all-nodes multicast address are not removed when
   the last IPv6 address is removed from an interface.

Reviewed by: bz, gnn

15 years agoDMA synchronization fixes:
delphij [Mon, 18 May 2009 01:51:52 +0000 (01:51 +0000)]
DMA synchronization fixes:

 - In bce_rx_intr(), use BUS_DMASYNC_POSTREAD instead of
   BUS_DMASYNC_POSTWRITE, as we want to "read" from the
   rx page chain pages.
 - Document why we need to do PREWRITE after we have updated
   the rx page chain pages.
 - In bce_intr(), use BUS_DMASYNC_POSTREAD and
   BUS_DMASYNC_PREREAD when before and after CPU "reading"
   the status block.
 - Adjust some nearby style mismatches/etc.

Pointed out by: yongari
Approved by: davidch (no objection) but bugs are mine :)

15 years agoRemove redundant whitespace
brian [Mon, 18 May 2009 01:45:52 +0000 (01:45 +0000)]
Remove redundant whitespace

15 years agoCorrect types of PHY, per http://bcm-specs.sipsolutions.net/PHYVersioning
imp [Mon, 18 May 2009 01:07:38 +0000 (01:07 +0000)]
Correct types of PHY, per http://bcm-specs.sipsolutions.net/PHYVersioning

# Note: The driver doesn't support either these PHY types, so this is
# effectively a nop.

Submitted by: "ddk"
Obtained from: http://paradox.lissyara.su/bwi.diff

15 years agotbr_timeout() is a timer driven function[1]. While the previous commit
bz [Mon, 18 May 2009 01:05:09 +0000 (01:05 +0000)]
tbr_timeout() is a timer driven function[1]. While the previous commit
made LINT happy this does the proper looping over all vnets as we are
only called `globally' and not once per vnet instance.

Reported by: zec [1]
Missed by: bz  [1] in r192264
Reviewed by: zec

15 years agoTurns out that my BCM4318 has a PCI ID of 0x4319, which lead me to
imp [Mon, 18 May 2009 01:00:11 +0000 (01:00 +0000)]
Turns out that my BCM4318 has a PCI ID of 0x4319, which lead me to
believe it was a BCM4319.  However, it is the a/b/g variation of the
BCM4318.  The chip itself is labelled BCM4318EKFBG, and the board is
BCM94318MKABG.

Paradox's patch includes the type of 802.11 wireless for each card,
but changes all the names (I don't think the latter is quite right).
Import that part of the patch, but keep the current set of BCM names
(with a minor tweak for the 4306 ones). I'll need to verify them via
some other means.

Obtained from: http://paradox.lissyara.su/bwi.diff (partially)

15 years agoSeveral changes to vfs_bio_clrbuf():
alc [Sun, 17 May 2009 23:25:53 +0000 (23:25 +0000)]
Several changes to vfs_bio_clrbuf():

Provide a more descriptive comment.

Eliminate dead code.  The page cannot possibly have PG_ZERO set.

Eliminate unnecessary blank lines.

Reviewed by: tegge

15 years agoDocument sbuf_new_auto().
brueffer [Sun, 17 May 2009 21:28:37 +0000 (21:28 +0000)]
Document sbuf_new_auto().

While here, add a missing `-' in phk's name.

MFC after: 3 days

15 years agoAdd a missing INIT_VNET_NET() to get VIMAGE closer to full LINT again.
bz [Sun, 17 May 2009 20:59:09 +0000 (20:59 +0000)]
Add a missing INIT_VNET_NET() to get VIMAGE closer to full LINT again.

15 years agoPut the structs under #ifndef VIMAGE_GLOBALS as some constants are not
bz [Sun, 17 May 2009 20:55:33 +0000 (20:55 +0000)]
Put the structs under #ifndef VIMAGE_GLOBALS as some constants are not
defined for VIMAGE_GLOBALS and thus broke the build for that option.

Reported by: csjp

15 years agoUnbreak options VIMAGE builds, in a followup to r192011 which did not
bz [Sun, 17 May 2009 20:53:10 +0000 (20:53 +0000)]
Unbreak options VIMAGE builds, in a followup to r192011 which did not
introduce INIT_VNET_NET() initializers necessary for accessing V_loif.

Submitted by: zec
Reviewed by: julian

15 years agoEliminate a pointless call to pmap_clear_reference() from vm_pageout_scan().
alc [Sun, 17 May 2009 20:40:41 +0000 (20:40 +0000)]
Eliminate a pointless call to pmap_clear_reference() from vm_pageout_scan().
If the page belongs to an object with a reference count of zero, then it
can't have any managed mappings on which to clear a reference bit.

15 years agoIntroduce vfs_bio_set_valid() and use it from ffs_realloccg(). This
alc [Sun, 17 May 2009 20:26:00 +0000 (20:26 +0000)]
Introduce vfs_bio_set_valid() and use it from ffs_realloccg().  This
eliminates the misuse of vfs_bio_clrbuf() by ffs_realloccg().

In collaboration with: tegge

15 years agoadd TRENDnet TEW-504UB/EU
sam [Sun, 17 May 2009 19:51:08 +0000 (19:51 +0000)]
add TRENDnet TEW-504UB/EU

Submitted by: Lucius Windschuh <lwindschuh@googlemail.com>

15 years agofix 11a channel use; mark OFDM operation correctly
sam [Sun, 17 May 2009 19:46:50 +0000 (19:46 +0000)]
fix 11a channel use; mark OFDM operation correctly

Submitted by: Lucius Windschuh <lwindschuh@googlemail.com>

15 years agoFix the acquisition of local locks via VOP_ADVLOCK() by the
rmacklem [Sun, 17 May 2009 19:33:48 +0000 (19:33 +0000)]
Fix the acquisition of local locks via VOP_ADVLOCK() by the
experimental nfsv4 server. It was setting the a_id argument
to a fixed value, but that wasn't sufficient for FreeBSD8.
Instead, set l_pid and l_sysid to 0 plus set the F_REMOTE
flag to indicate that these fields are used to check for
same lock owner. Since, for NFSv4, a lockowner is a ClientID plus
an up to 1024byte name, it can't be put in l_sysid easily.
I also renamed the p variable to td, since it's a thread ptr.

Approved by: kib (mentor)

15 years agoAdded a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of
rmacklem [Sun, 17 May 2009 17:54:01 +0000 (17:54 +0000)]
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of
nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty
descriptor strings and reformatted some overly long lines.

Approved by: kib (mentor)

15 years agoPrint an extra newline when not at the first column already.
ed [Sun, 17 May 2009 16:17:48 +0000 (16:17 +0000)]
Print an extra newline when not at the first column already.

This makes siginfo output look a lot better when pressing it the first
time when in sh(1), for example:

$ load: 0.00  cmd: sh 1945 [ttyin] 3.94r 0.00u 0.00s 0% 1960k
load: 0.00  cmd: sh 1945 [ttyin] 4.19r 0.00u 0.00s 0% 1960k

will now become:

$
load: 0.00  cmd: sh 1945 [ttyin] 3.94r 0.00u 0.00s 0% 1960k
load: 0.00  cmd: sh 1945 [ttyin] 4.19r 0.00u 0.00s 0% 1960k

15 years agoSeveral cleanups to tty_info(), better known as Ctrl-T.
ed [Sun, 17 May 2009 12:30:25 +0000 (12:30 +0000)]
Several cleanups to tty_info(), better known as Ctrl-T.

- Only pick up PROC_LOCK once, which means we can drop the PGRP_LOCK
  right after picking up PROC_LOCK for the first time.

- Print the process real time, making it consistent with tools like
  time(1).

- Use `p' and `td' to reference the process/thread we are going to
  print. Only use pick-variables inside the loops. We already did this
  for the threads, but not the processes.

15 years agoFix the example.
trasz [Sun, 17 May 2009 12:21:11 +0000 (12:21 +0000)]
Fix the example.

15 years agoImprove the accf_dns_load description.
brueffer [Sun, 17 May 2009 10:58:50 +0000 (10:58 +0000)]
Improve the accf_dns_load description.

15 years ago- do not create and mount new file systems on top of the old ones on every
danger [Sun, 17 May 2009 08:25:02 +0000 (08:25 +0000)]
- do not create and mount new file systems on top of the old ones on every
  invocation of this script once we already have one
  (in case tmpmfs="YES").

Reviewed by: dougb

15 years agoMerge r191964: Eliminate a case of unnecessary page queues locking.
alc [Sun, 17 May 2009 06:45:30 +0000 (06:45 +0000)]
Merge r191964: Eliminate a case of unnecessary page queues locking.

15 years agoadd zfs oid to bsnmpd
kmacy [Sun, 17 May 2009 05:54:25 +0000 (05:54 +0000)]
add zfs oid to bsnmpd

PR: bin/129360
Submitted by: Ulrich Spoerlein

15 years agoWhen finding processes, ignore ourself and our ancestors. It is almost
brian [Sun, 17 May 2009 04:34:14 +0000 (04:34 +0000)]
When finding processes, ignore ourself and our ancestors.  It is almost
always surprising when you kill a 'sh -c ...' ancestor or when you kill
yourself when using -f.

Add a -a switch for backwards compatibility.

MFC after: 3 weeks

15 years agoset createtxg prop name
kmacy [Sun, 17 May 2009 04:04:25 +0000 (04:04 +0000)]
set createtxg prop name

PR: bin/130105

15 years agoAdd a -d option to ps to display descendant info with the output.
brian [Sun, 17 May 2009 04:00:43 +0000 (04:00 +0000)]
Add a -d option to ps to display descendant info with the output.
This is similar to linux's -H (or -f) switch.

MFC after: 3 weeks

15 years agoSAVESTART implies SAVENAME
kmacy [Sun, 17 May 2009 01:31:28 +0000 (01:31 +0000)]
SAVESTART implies SAVENAME

15 years agoenable adaptive spinning on zfs locks
kmacy [Sat, 16 May 2009 23:56:45 +0000 (23:56 +0000)]
enable adaptive spinning on zfs locks

15 years agoChanged sys/fs/nfs_clbio.c in the same way Alan Cox changed
rmacklem [Sat, 16 May 2009 22:31:38 +0000 (22:31 +0000)]
Changed sys/fs/nfs_clbio.c in the same way Alan Cox changed
sys/nfsclient/nfs_bio.c for r192134, so that the sources stay
in sync.

Approved by: kib (mentor)

15 years agoEnhance the safety of the -U option:
dougb [Sat, 16 May 2009 22:22:31 +0000 (22:22 +0000)]
Enhance the safety of the -U option:
1. In several places make sure that the mtree database is not empty using
the -s argument to test instead of -f. (I thought I'd already changed this,
but obviously not.)
2. When deriving the list of changed files use a colon-delimited list.
3. If the list of changed files is empty, unset the variable, and test
for a non-empty CHANGED in diff_loop() before entering that routine.

Enhance the speed of the -U option by using an internal case statement
instead of echoing the variable to an external grep for every file.

Fix indentation on one line.

15 years agoDrop clause 3 of the license as per rev. 1.35 from OpenBSD.
brueffer [Sat, 16 May 2009 22:18:58 +0000 (22:18 +0000)]
Drop clause 3 of the license as per rev. 1.35 from OpenBSD.

Obtained from: OpenBSD
MFC after: 3 days

15 years agocorrect range in comment
kmacy [Sat, 16 May 2009 22:08:00 +0000 (22:08 +0000)]
correct range in comment
pointed out by alc

15 years agoupdate vm map comment
kmacy [Sat, 16 May 2009 22:00:13 +0000 (22:00 +0000)]
update vm map comment

pointed out by Larry Rosenman

15 years agoSync up to rev. 1.25 from OpenBSD:
brueffer [Sat, 16 May 2009 21:38:55 +0000 (21:38 +0000)]
Sync up to rev. 1.25 from OpenBSD:

1.19: nop
1.20: nop
1.21: nop
1.22: typo fixed
1.23: license clauses 3 and 4 nuked
1.24: nop
1.25: Sun Crypto Accelerator 1000 has a 5821 chip on it

Obtained from: OpenBSD
MFC after: 3 days

15 years agoAdd -L to the mtree invocation to chase symbolic links that are present
dougb [Sat, 16 May 2009 21:32:09 +0000 (21:32 +0000)]
Add -L to the mtree invocation to chase symbolic links that are present
in "Service jails."

Submitted by: Jeremie Le Hen <jeremie@le-hen.org>

15 years agoIncrease default kernel map to 512GB
kmacy [Sat, 16 May 2009 20:57:08 +0000 (20:57 +0000)]
Increase default kernel map to 512GB

I briefly discussed this with alc. It could lead to problems for greater than 64GB.
However, that seems unlikely in practice.

15 years ago1. New feature; option to have the script loop until a specified hostname
dougb [Sat, 16 May 2009 20:55:28 +0000 (20:55 +0000)]
1. New feature; option to have the script loop until a specified hostname
(localhost by default) can be successfully looked up. Off by default.
2. New feature: option to create a forwarder configuration file based on
the contents of /etc/resolv.conf. This allows you to utilize a local
resolver for better performance, less network traffic, custom zones, etc.
while still relying on the benefits of your local network resolver.
Off by default.
3. Add named-checkconf into the startup routine. This will prevent named
from trying to start in a situation where it would not be possible to do
so.

15 years ago- allow forced unmounts
kmacy [Sat, 16 May 2009 20:33:13 +0000 (20:33 +0000)]
- allow forced unmounts
- don't assume snapshot was auto-mounted

15 years agoTrim trailing whitespace from the end of a line
dougb [Sat, 16 May 2009 20:26:01 +0000 (20:26 +0000)]
Trim trailing whitespace from the end of a line

15 years agoonly use direct map if system has more than 2GB
kmacy [Sat, 16 May 2009 20:09:07 +0000 (20:09 +0000)]
only use direct map if system has more than 2GB

15 years agoapply band-aid to x86_64 systems with more physical memory than kmem by allocating...
kmacy [Sat, 16 May 2009 19:17:15 +0000 (19:17 +0000)]
apply band-aid to x86_64 systems with more physical memory than kmem by allocating from the direct map

15 years agoSomewhere between 2.6.23 and 2.6.27, Linux added SOCK_CLOEXEC and
dchagin [Sat, 16 May 2009 18:48:41 +0000 (18:48 +0000)]
Somewhere between 2.6.23 and 2.6.27, Linux added SOCK_CLOEXEC and
SOCK_NONBLOCK flags, that allow to save fcntl() calls.

Implement a variation of the socket() syscall which takes a flags
in addition to the type argument.

Approved by: kib (mentor)
MFC after: 1 month

15 years agoReturn EINVAL in case when the incorrect or unsupported
dchagin [Sat, 16 May 2009 18:46:51 +0000 (18:46 +0000)]
Return EINVAL in case when the incorrect or unsupported
type argument is specified.

Do not map type argument value as its Linux values are
identical to FreeBSD values.

Approved by: kib (mentor)

15 years agoUse the protocol family constants for the domain argument validation.
dchagin [Sat, 16 May 2009 18:44:56 +0000 (18:44 +0000)]
Use the protocol family constants for the domain argument validation.
Return immediately when the socket() failed.

Approved by: kib (mentor)
MFC after: 1 month

15 years agoEmulate SO_PEERCRED socket option.
dchagin [Sat, 16 May 2009 18:42:18 +0000 (18:42 +0000)]
Emulate SO_PEERCRED socket option.
Temporarily use 0 for pid member as the FreeBSD does not cache remote
UNIX domain socket peer pid.

PR: kern/102956
Reviewed by: rwatson
Approved by: kib (mentor)
MFC after: 1 month

15 years ago- Fix spelling.
lulf [Sat, 16 May 2009 15:21:08 +0000 (15:21 +0000)]
- Fix spelling.

15 years agoo Add missed semicolon in action script.
maxim [Sat, 16 May 2009 15:12:56 +0000 (15:12 +0000)]
o Add missed semicolon in action script.

PR: conf/134579
Submitted by: Lucius Windschuh
MFC after: 1 week

15 years agoAdd bwi(4) and urtw(4).
brueffer [Sat, 16 May 2009 12:12:31 +0000 (12:12 +0000)]
Add bwi(4) and urtw(4).

15 years agoMake the HARDWARE section ready for the automatic hardware notes generation,
brueffer [Sat, 16 May 2009 12:11:06 +0000 (12:11 +0000)]
Make the HARDWARE section ready for the automatic hardware notes generation,
fix the date format.

15 years agoAdd support for booting from raidz1 and raidz2 pools.
dfr [Sat, 16 May 2009 10:48:20 +0000 (10:48 +0000)]
Add support for booting from raidz1 and raidz2 pools.

15 years agoAdd a manpage for the bwi(4) driver.
brueffer [Sat, 16 May 2009 10:42:00 +0000 (10:42 +0000)]
Add a manpage for the bwi(4) driver.

15 years agoFixed the Null callback RPCs so that they work with the new krpc. This
rmacklem [Sat, 16 May 2009 03:12:55 +0000 (03:12 +0000)]
Fixed the Null callback RPCs so that they work with the new krpc. This
required two changes: setting the program and version numbers before
connect and fixing the handling of the Null Rpc case in newnfs_request().

Approved by: kib (mentor)

15 years agoRemove do-nothing code that was required to dirty the old buffer on Alpha.
des [Fri, 15 May 2009 21:34:58 +0000 (21:34 +0000)]
Remove do-nothing code that was required to dirty the old buffer on Alpha.

Coverity ID: 838
Approved by: jhb, alc

15 years agoAdd a missing break in a switch statement.
brueffer [Fri, 15 May 2009 21:06:28 +0000 (21:06 +0000)]
Add a missing break in a switch statement.

Found with: Coverity Prevent(tm)
CID: 4302
MFC after: 2 weeks

15 years agoReduce diff against my local version: replace malloc+memset() cases to calloc().
delphij [Fri, 15 May 2009 19:41:10 +0000 (19:41 +0000)]
Reduce diff against my local version: replace malloc+memset() cases to calloc().

15 years agoMove the nfsstat structure and proc/op number definitions on the
rmacklem [Fri, 15 May 2009 19:33:59 +0000 (19:33 +0000)]
Move the nfsstat structure and proc/op number definitions on the
experimental nfs subsystem from sys/fs/nfs/nfs.h and sys/fs/nfs/nfsproto.h
to sys/fs/nfs/nfsport.h and rename nfsstat to ext_nfsstat. This was done
so that src/usr.bin/nfsstat.c could use it alongside the regular nfs
include files and struct nfsstat.

Approved by: kib (mentor)

15 years agoDevfs replaces file ops vector with devfs-specific one in devfs_open(),
kib [Fri, 15 May 2009 19:23:05 +0000 (19:23 +0000)]
Devfs replaces file ops vector with devfs-specific one in devfs_open(),
before the struct file is fully initialized in vn_open(), in particular,
fp->f_vnode is NULL. Other thread calling file operation before f_vnode
is set results in NULL pointer dereference in devvn_refthread().

Initialize f_vnode before calling d_fdopen() cdevsw method, that might
set file ops too.

Reported and tested by: Chris Timmons <cwt networks cwu edu>
(RELENG_7 version)
MFC after: 3 days

15 years agoUpdate the KVM backend for malloc stats to catch up to the internal structure
jhb [Fri, 15 May 2009 18:25:44 +0000 (18:25 +0000)]
Update the KVM backend for malloc stats to catch up to the internal structure
BI change from the addition of DTrace malloc(9) probes.

Submitted by: Ben Kelly  ben of wanderview dot com

15 years agoThe module name convention is foo, not if_foo.
imp [Fri, 15 May 2009 17:02:11 +0000 (17:02 +0000)]
The module name convention is foo, not if_foo.

15 years agoCleanup module declarations:
imp [Fri, 15 May 2009 16:38:42 +0000 (16:38 +0000)]
Cleanup module declarations:
Use "bwi" consistently
No need to have separate cardbus attachment
No need to declare a module version

15 years agoModify the diskless booting code in sys/fs/nfsclient to be compatible
rmacklem [Fri, 15 May 2009 16:03:11 +0000 (16:03 +0000)]
Modify the diskless booting code in sys/fs/nfsclient to be compatible
with what is in sys/nfsclient, so that it will at least build now.

Approved by: kib (mentor)

15 years agoRevert r192094. The revision caused problems for sysctl(3) consumers
kib [Fri, 15 May 2009 14:41:44 +0000 (14:41 +0000)]
Revert r192094. The revision caused problems for sysctl(3) consumers
that expect that oldlen is filled with required buffer length even when
supplied buffer is too short and returned error is ENOMEM.

Redo the fix for kern.proc.filedesc, by reverting the req->oldidx when
remaining buffer space is too short for the current kinfo_file structure.
Also, only ignore ENOMEM. We have to convert ENOMEM to no error condition
to keep existing interface for the sysctl, though.

Reported by: ed, Florian Smeets <flo kasimir com>
Tested by: pho

15 years agoTurn consolectl into a simple device node, not a TTY.
ed [Fri, 15 May 2009 14:30:37 +0000 (14:30 +0000)]
Turn consolectl into a simple device node, not a TTY.

Apart from the 16 virtual terminals, Syscons allocates two device nodes
that should not really be TTYs, even though they are. One of them is
consolectl. In RELENG_7 and before, these device nodes are used in
single user mode. After I simplified input path, we only use this device
node to call ioctl() on (moused, Xorg, vidcontrol).

When you call ioctl() on consolectl, it will behave the same as being
called on the first window.

15 years agoThis driver can be loaded as a module now.
brueffer [Fri, 15 May 2009 13:26:54 +0000 (13:26 +0000)]
This driver can be loaded as a module now.

MFC after: 3 days

15 years agoXref missing wlan drivers.
brueffer [Fri, 15 May 2009 10:11:54 +0000 (10:11 +0000)]
Xref missing wlan drivers.

15 years agoSort SEE ALSO section, remove trailing dot.
brueffer [Fri, 15 May 2009 10:06:33 +0000 (10:06 +0000)]
Sort SEE ALSO section, remove trailing dot.

15 years agoFix name for driver to assign to the SPI device.
imp [Fri, 15 May 2009 04:49:20 +0000 (04:49 +0000)]
Fix name for driver to assign to the SPI device.

15 years agoEliminate unnecessary clearing of the page's dirty mask from various
alc [Fri, 15 May 2009 04:33:35 +0000 (04:33 +0000)]
Eliminate unnecessary clearing of the page's dirty mask from various
getpages functions.

Eliminate a stale comment.

15 years agoAllow a comma-separated list of network interfaces to be specified via the
cperciva [Fri, 15 May 2009 00:18:31 +0000 (00:18 +0000)]
Allow a comma-separated list of network interfaces to be specified via the
netDev option in install.cfg.

Submitted by: randi
MFC after: 1 week

15 years agoAs the comment says, close() frees the variable, record. So we obtain
delphij [Thu, 14 May 2009 23:09:33 +0000 (23:09 +0000)]
As the comment says, close() frees the variable, record.  So we obtain
the length by evaluating the value from the copy, cbuf instead.  This
fixes a crash caused by previous commit (use-after-free)

Submitted by: Dimitry Andric <dimitry andric com>
Pointy hat to: delphij

15 years agoSome comment/space changes (FALLTHRU -> FALLTHROUGH, space after while).
delphij [Thu, 14 May 2009 22:36:56 +0000 (22:36 +0000)]
Some comment/space changes (FALLTHRU -> FALLTHROUGH, space after while).

15 years agoTry to workaround a race where bge_stop() may sneak in when bge_rxeof()
delphij [Thu, 14 May 2009 22:33:37 +0000 (22:33 +0000)]
Try to workaround a race where bge_stop() may sneak in when bge_rxeof()
drops and re-grabs the softc mutex in the middle, resulting in kernel
trap 12.  This may happen when a lot of traffic is being hammered on
one bge(4) interface while the system is shutting down.

Reported by: Alexander Sack <pisymbol gmail com>
PR: kern/134548
MFC After: 2 weeks

15 years ago- Use a separate sx lock to try to limit the number of concurrent userland
jhb [Thu, 14 May 2009 22:01:32 +0000 (22:01 +0000)]
- Use a separate sx lock to try to limit the number of concurrent userland
  sysctl requests to avoid wiring too much user memory.  Only grab this
  lock if the user's old buffer is larger than a page as a tradeoff to
  allow more concurrency for common small requests.
- Just use a shared lock on the sysctl tree for user sysctl requests now.

MFC after: 1 week

15 years agoTrim the default set of device hints on i386 and amd64:
jhb [Thu, 14 May 2009 21:53:35 +0000 (21:53 +0000)]
Trim the default set of device hints on i386 and amd64:
- Remove vga0 and the disabled uart2/uart3 hints from both platforms.
- Remove hints for ISA adv0, bt0, aha0, aic0, ed0, cs0, sn0, ie0, fe0, and
  le0 from i386.  All these hints were marked 'disabled' and thus already
  did not work "out of the box".

Discussed with: imp

15 years agoApply changes to the experimental nfs server so that it uses the security
rmacklem [Thu, 14 May 2009 21:39:08 +0000 (21:39 +0000)]
Apply changes to the experimental nfs server so that it uses the security
flavors as exported in FreeBSD-CURRENT. This allows it to use a
slightly modified mountd.c instead of a different utility.

Approved by: kib (mentor)

15 years agoStaticize two functions not used outside of in_pcb.c: in_pcbremlists() and
rwatson [Thu, 14 May 2009 20:59:36 +0000 (20:59 +0000)]
Staticize two functions not used outside of in_pcb.c: in_pcbremlists() and
db_print_inpcb().

MFC after: 1 month

15 years agoChange the file names in the comments in sys/fs/nfs/nfs_var.h so
rmacklem [Thu, 14 May 2009 20:39:09 +0000 (20:39 +0000)]
Change the file names in the comments in sys/fs/nfs/nfs_var.h so
that they are the names used in FreeBSD-CURRENT. Also shuffled a
few entries around, so that they under the correct comment.

Approved by: kib (mentor)

15 years agoFreeBSD right now support 32 CPUs on all the architectures at least.
attilio [Thu, 14 May 2009 17:43:00 +0000 (17:43 +0000)]
FreeBSD right now support 32 CPUs on all the architectures at least.
With the arrival of 128+ cores it is necessary to handle more than that.
One of the first thing to change is the support for cpumask_t that needs
to handle more than 32 bits masking (which happens now).  Some places,
however, still assume that cpumask_t is a 32 bits mask.
Fix that situation by using always correctly cpumask_t when needed.

While here, remove the part under STOP_NMI for the Xen support as it
is broken in any case.

Additively make ipi_nmi_pending as static.

Reviewed by: jhb, kmacy
Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>

15 years agoAvoid floating point arithmetic while calculating iquiry length.
emax [Thu, 14 May 2009 17:10:19 +0000 (17:10 +0000)]
Avoid floating point arithmetic while calculating iquiry length.

Submitted by: Iain Hibbert < plunky -at- rya-online -dot- net >
MFC after: 1 week

15 years agoImprove style(9)
raj [Thu, 14 May 2009 16:56:56 +0000 (16:56 +0000)]
Improve style(9)

15 years agoPowerPC common SMP startup and time base rework.
raj [Thu, 14 May 2009 16:48:25 +0000 (16:48 +0000)]
PowerPC common SMP startup and time base rework.

- make mftb() shared, rewrite in C, provide complementary mttb()
- adjust SMP startup per the above, additional comments, minor naming
  changes
- eliminate redundant TB defines, other minor cosmetics

Reviewed by: marcel, nwhitehorn
Obtained from: Freescale, Semihalf

15 years agoadd missing type for SYSCTL_PROC items; w/o a type you can view the value
sam [Thu, 14 May 2009 16:25:57 +0000 (16:25 +0000)]
add missing type for SYSCTL_PROC items; w/o a type you can view the value
but not change it

15 years agocorrect handling of ctl frames: the sender's address is always i_addr2 for
sam [Thu, 14 May 2009 16:23:24 +0000 (16:23 +0000)]
correct handling of ctl frames: the sender's address is always i_addr2 for
frames we should expect to process (old code was trying to handle frames
we should never see--like ACK)

Reviewed by: thompsa, cbzimmer

15 years agoMFi386: revision 192050
nyan [Thu, 14 May 2009 16:01:29 +0000 (16:01 +0000)]
MFi386: revision 192050

  Implement simple machine check support.

15 years ago- Add a void pointer to the ata-pci controller softc to allow
jhb [Thu, 14 May 2009 14:57:13 +0000 (14:57 +0000)]
- Add a void pointer to the ata-pci controller softc to allow
  chipset-specific code to attach chipset-specific data.
- Use chipset-specific data in the acard and promise chipsets rather than
  changing the ivars of ATA PCI devices.  ivars are reserved for use by the
  parent bus driver and are _not_ available for use by devices directly.
  This fixes a panic during sysctl -a with certain Promise controllers with
  ACPI enabled.

Reviewed by: mav
Tested by: Magnus Kling (kingfon @ gmail) (on 7)
MFC after: 3 days

15 years agoCompare the correct variable against NULL.
brueffer [Thu, 14 May 2009 13:32:33 +0000 (13:32 +0000)]
Compare the correct variable against NULL.

Reviewed by: scottl
Found with: Coverity Prevent(tm)
CID: 821
MFC after: 2 weeks

15 years agoRemove usb_ethersubr.c missed reference -- we don't build ousb
rwatson [Thu, 14 May 2009 12:39:22 +0000 (12:39 +0000)]
Remove usb_ethersubr.c missed reference -- we don't build ousb
with LINT, so the tinderbox didn't pick this up.