]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years ago- rename atomic.S and crc32.c to avoid collisions when linking zfs in to the kernel
Kip Macy [Sat, 9 May 2009 01:45:55 +0000 (01:45 +0000)]
- rename atomic.S and crc32.c to avoid collisions when linking zfs in to the kernel
- update Makefile
- ifdef out acl_{alloc, free}, they aren't used by zfs and conflict with existing in-kernel routines

15 years agoUpdate to less v429.
Xin LI [Sat, 9 May 2009 01:35:27 +0000 (01:35 +0000)]
Update to less v429.

15 years agoVendor import of less v429
Xin LI [Fri, 8 May 2009 23:45:27 +0000 (23:45 +0000)]
Vendor import of less v429

15 years agoFlatten all tags of the dist tree of less.
Xin LI [Fri, 8 May 2009 23:34:35 +0000 (23:34 +0000)]
Flatten all tags of the dist tree of less.

15 years ago- Fixed incorrect packet length problem caused be earlier change to
David Christensen [Fri, 8 May 2009 22:20:45 +0000 (22:20 +0000)]
- Fixed incorrect packet length problem caused be earlier change to
  support ZERO_COPY_SOCKETS.
- Created #define for context initialization retry count.

MFC after: 1 week

15 years agoRegenerate system call tables to use SVN ids.
Ed Schouten [Fri, 8 May 2009 20:16:04 +0000 (20:16 +0000)]
Regenerate system call tables to use SVN ids.

15 years agoRegenerate ibcs2 system call table.
Ed Schouten [Fri, 8 May 2009 20:08:43 +0000 (20:08 +0000)]
Regenerate ibcs2 system call table.

15 years agoBurn TTY ioctl bridges in compat layers.
Ed Schouten [Fri, 8 May 2009 20:06:37 +0000 (20:06 +0000)]
Burn TTY ioctl bridges in compat layers.

I really don't want any pieces of code to include ioctl_compat.h, so let
the ibcs2 and svr4 compat leave sgtty alone. If they want to support
sgtty, they should emulate it on top of termios, not sgtty.

The code has been marked with BURN_BRIDGES for a long time. ibcs2 and
svr4 are not really popular pieces of code anyway.

15 years agoUse log(9) for debug and status messages and hide some of the details with
Doug Rabson [Fri, 8 May 2009 15:14:52 +0000 (15:14 +0000)]
Use log(9) for debug and status messages and hide some of the details with
macros to allow for future flexibility in logging.

Submitted by: zachary dot loafman at isilon dot com

15 years agoA NOP change: style / whitespace cleanup of the noise that slipped
Marko Zec [Fri, 8 May 2009 14:34:25 +0000 (14:34 +0000)]
A NOP change: style / whitespace cleanup of the noise that slipped
into r191816.

Spotted by: bz
Approved by: julian (mentor) (an earlier version of the diff)

15 years agoRemove a bogus check that unintentionally slipped in r191816.
Marko Zec [Fri, 8 May 2009 14:28:06 +0000 (14:28 +0000)]
Remove a bogus check that unintentionally slipped in r191816.

This change has no functional impact on nooptions VIMAGE builds.
Submitted by: bz

15 years agoIntroduce a new virtualization container, provisionally named vprocg, to hold
Marko Zec [Fri, 8 May 2009 14:11:06 +0000 (14:11 +0000)]
Introduce a new virtualization container, provisionally named vprocg, to hold
virtualized instances of hostname and domainname, as well as a new top-level
virtualization struct vimage, which holds pointers to struct vnet and struct
vprocg.  Struct vprocg is likely to become replaced in the near future with
a new jail management API import.

As a consequence of this change, change struct ucred to point to a struct
vimage, instead of directly pointing to a vnet.

Merge vnet / vimage / ucred refcounting infrastructure from p4 / vimage
branch.

Permit kldload / kldunload operations to be executed only from the default
vimage context.

This change should have no functional impact on nooptions VIMAGE kernel
builds.

Reviewed by: bz
Approved by: julian (mentor)

15 years agoSync ubsec(4) with reality, note BCM5825 support.
Philip Paeps [Fri, 8 May 2009 14:09:31 +0000 (14:09 +0000)]
Sync ubsec(4) with reality, note BCM5825 support.

Pointed out by: brueffer
MFC after: 1 minute

15 years agoDrain the tasks before the interface stop call in case a restart was queued.
Andrew Thompson [Fri, 8 May 2009 13:44:33 +0000 (13:44 +0000)]
Drain the tasks before the interface stop call in case a restart was queued.

15 years agoAdd one more board ID with inverted external amplifier control.
Alexander Motin [Fri, 8 May 2009 03:19:57 +0000 (03:19 +0000)]
Add one more board ID with inverted external amplifier control.

PR: kern/121156

15 years agoChange the internal buffer used to store input lines from a static buffer
Stephane E. Potvin [Fri, 8 May 2009 02:18:46 +0000 (02:18 +0000)]
Change the internal buffer used to store input lines from a static buffer
to a dynamically allocated one in order to support input lines of
arbitrary length.

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

15 years agokill more portability functions that are no longer useful
Sam Leffler [Fri, 8 May 2009 00:23:00 +0000 (00:23 +0000)]
kill more portability functions that are no longer useful

15 years agokill unused OS_GETUPTIME
Sam Leffler [Fri, 8 May 2009 00:15:27 +0000 (00:15 +0000)]
kill unused OS_GETUPTIME

15 years agodon't call vn_rele_async_fini in the !_KERNEL case
Kip Macy [Thu, 7 May 2009 23:34:41 +0000 (23:34 +0000)]
don't call vn_rele_async_fini in the !_KERNEL case

15 years agodefine VN_RELE_ASYNC for use by libzpool
Kip Macy [Thu, 7 May 2009 23:02:51 +0000 (23:02 +0000)]
define VN_RELE_ASYNC for use by libzpool

15 years agomove VN_RELE_ASYNC to the compatibility layer with the rest of the VN_* defines
Kip Macy [Thu, 7 May 2009 23:02:15 +0000 (23:02 +0000)]
move VN_RELE_ASYNC to the compatibility layer with the rest of the VN_* defines

15 years agoPartially revert r191171, which went too far in trying
Tim Kientzle [Thu, 7 May 2009 23:01:03 +0000 (23:01 +0000)]
Partially revert r191171, which went too far in trying
to eliminate some duplicated code.  In particular,
archive_read_open_filename() has different close
handling than archive_read_open_fd(), so delegating
the former to the latter in the degenerate case
(a NULL filename is treated as stdin) broke reading
from pipelines.  In particular, this fixes occasional
port failures that were seen when using "gunzip | tar"
pipelines under /bin/csh.

Thanks to Alexey Shuvaev for reporting this failure and
patiently helping me to track down the cause.

15 years agoavoid LOR and gratuitous extra lock acquisitions by moving user_evict list buffers to
Kip Macy [Thu, 7 May 2009 21:51:13 +0000 (21:51 +0000)]
avoid LOR and gratuitous extra lock acquisitions by moving user_evict list buffers to
a temporary list

15 years agoAllow the VM to provide backpressure on the ARC cache as it does
Kip Macy [Thu, 7 May 2009 20:57:06 +0000 (20:57 +0000)]
Allow the VM to provide backpressure on the ARC cache as it does
on Solaris.

15 years agoAsynchronously release vnodes to avoid blocking on range locks when calling back...
Kip Macy [Thu, 7 May 2009 20:28:06 +0000 (20:28 +0000)]
Asynchronously release vnodes to avoid blocking on range locks when calling back in to zfs.
This is based on a fix that went in to opensolaris on March 9th. However, it uses a dedicated
thread instead of a Solaris' taskq to avoid doing a blocking memory allocation with the vnode
interlock held.

This fixes a long-time deadlock in ZFS. This is not, strictly speaking, an LOR. The spa_zio
thread releases a vnode, this calls in to vn_reclaim which in turn needs to acquire range locks
to sync dirty data out to disk. The range locks are already held by a user-level process waiting
on a condition variable that it the process is waiting on a spa_zio thread to signal it on. The
process could not be signalled because the spa_zio thread could not proceed.

The nature of this problem was not apparent due to ZFS locks opting out of witness which meant
that DDB did not know about the locks that were held by ZFS.

Reviewed by: pjd
MFC after: 7 days

15 years agoNo man page currently exists so comment the two uncommented
Kip Macy [Thu, 7 May 2009 19:57:14 +0000 (19:57 +0000)]
No man page currently exists so comment the two uncommented
non-trivial functions

15 years agoGive vfs_getopt the type it's expecting.
Jamie Gritton [Thu, 7 May 2009 19:46:29 +0000 (19:46 +0000)]
Give vfs_getopt the type it's expecting.
Write 100 times: "32 bits is so twentieth century."

Noticed by: dchagin

15 years agoAdd ID of one more SII3132 revision found on adaptec aar-1225sa rev a2.
Alexander Motin [Thu, 7 May 2009 19:17:06 +0000 (19:17 +0000)]
Add ID of one more SII3132 revision found on adaptec aar-1225sa rev a2.

PR: kern/127289

15 years agoMove the per-prison Linux MIB from a private one-off pointer to the new
Jamie Gritton [Thu, 7 May 2009 18:36:47 +0000 (18:36 +0000)]
Move the per-prison Linux MIB from a private one-off pointer to the new
OSD-based jail extensions.  This allows the Linux MIB to accessed via
jail_set and jail_get, and serves as a demonstration of adding jail support
to a module.

Reviewed by: dchagin, kib
Approved by: bz (mentor)

15 years agoEliminate the loop and the call to pause(9) in vfs_vget_ino(). If
Konstantin Belousov [Thu, 7 May 2009 18:14:21 +0000 (18:14 +0000)]
Eliminate the loop and the call to pause(9) in vfs_vget_ino(). If
vfs_busy(MBF_NOWAIT) failed, unlock the vnode and sleep in vfs_busy().

Suggested and reviewed by: jeff
Tested by: pho
MFC after: 3 weeks

15 years agoAdd PCI IDs for the Broadcom 5825 incarnation.
Philip Paeps [Thu, 7 May 2009 18:03:47 +0000 (18:03 +0000)]
Add PCI IDs for the Broadcom 5825 incarnation.

Submitted by: Brian A. Seklecki <bseklecki -at- collaborativefusion.com>
MFC after: 1 day

15 years agoIf we have a regular rint handler, never go into rint_bypass mode.
Ed Schouten [Thu, 7 May 2009 17:39:23 +0000 (17:39 +0000)]
If we have a regular rint handler, never go into rint_bypass mode.

It turns out if we called cfmakeraw() on a TTY with only a rint handler
in place, it could inject data into the TTY, even though it should be
redirected. Always take a look at the hooks before looking at the
termios flags.

15 years agoUse a 32 bit type for the interface mask as this equals the max interface
Andrew Thompson [Thu, 7 May 2009 17:05:03 +0000 (17:05 +0000)]
Use a 32 bit type for the interface mask as this equals the max interface
count.

15 years agorepository sync to multi-OS repo ... spaceing change
Randall Stewart [Thu, 7 May 2009 16:43:49 +0000 (16:43 +0000)]
repository sync to multi-OS repo ... spaceing change

15 years agoABI expansions to hopefully future-proof our MIB/netstat code for 8.0
Randall Stewart [Thu, 7 May 2009 16:42:45 +0000 (16:42 +0000)]
ABI expansions to hopefully future-proof our MIB/netstat code for 8.0

15 years agoUpdate example with feedback from thompsa@.
Ed Maste [Thu, 7 May 2009 16:31:50 +0000 (16:31 +0000)]
Update example with feedback from thompsa@.

15 years agoAdd KTR(9) tracing for futex emulation.
Dmitry Chagin [Thu, 7 May 2009 16:14:31 +0000 (16:14 +0000)]
Add KTR(9) tracing for futex emulation.

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

15 years agoAdd my birthday
Steven Kreuzer [Thu, 7 May 2009 15:32:13 +0000 (15:32 +0000)]
Add my birthday

Approved by: wxs (mentor)

15 years agoAdding myself as a ports committer
Steven Kreuzer [Thu, 7 May 2009 14:39:01 +0000 (14:39 +0000)]
Adding myself as a ports committer

Approved by: wxs (mentor)

15 years agos/ath0/wlan0/ since we no longer use the real device directly.
Ed Maste [Thu, 7 May 2009 14:34:49 +0000 (14:34 +0000)]
s/ath0/wlan0/ since we no longer use the real device directly.

15 years agoLinux exports HZ value to user space via AT_CLKTCK auxiliary vector entry,
Dmitry Chagin [Thu, 7 May 2009 14:24:50 +0000 (14:24 +0000)]
Linux exports HZ value to user space via AT_CLKTCK auxiliary vector entry,
which is available for Glibc as sysconf(_SC_CLK_TCK). If AT_CLKTCK entry is
not exported, Glibc uses 100.

linux_times() shall use the value that is exported to user space.

Pointyhat to: dchagin

PR: kern/134251
Approved by: kib (mentor)
MFC after: 2 weeks

15 years agoAdd tcsetsid(3).
Ed Schouten [Thu, 7 May 2009 13:49:48 +0000 (13:49 +0000)]
Add tcsetsid(3).

The entire world seems to use the non-standard TIOCSCTTY ioctl to make a
TTY a controlling terminal of a session. Even though tcsetsid(3) is also
non-standard, I think it's a lot better to use in our own source code,
mainly because it's similar to tcsetpgrp(), tcgetpgrp() and tcgetsid().

I stole the idea from QNX. They do it the other way around; their
TIOCSCTTY is just a wrapper around tcsetsid(). tcsetsid() then calls
into an IPC framework.

15 years agoChange linux struct tms definition to match actual linux one.
Dmitry Chagin [Thu, 7 May 2009 12:55:58 +0000 (12:55 +0000)]
Change linux struct tms definition to match actual linux one.

Approved by: kib (mentor)
MFC after: 2 weeks

15 years agoAdd preliminary KTR(9) support to the linux emulation layer.
Dmitry Chagin [Thu, 7 May 2009 10:01:05 +0000 (10:01 +0000)]
Add preliminary KTR(9) support to the linux emulation layer.

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

15 years agoTo avoid excessive code duplication move MI definitions to the MI
Dmitry Chagin [Thu, 7 May 2009 09:39:20 +0000 (09:39 +0000)]
To avoid excessive code duplication move MI definitions to the MI
header file. As it is defined in Linux.

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

15 years agoReturn EAFNOSUPPORT instead of EINVAL in case when the incorrect or
Dmitry Chagin [Thu, 7 May 2009 09:34:02 +0000 (09:34 +0000)]
Return EAFNOSUPPORT instead of EINVAL in case when the incorrect or
unsupported domain argument is specified.

Approved by: kib (mentor)

15 years agoEliminate an incorrect comment.
Alan Cox [Thu, 7 May 2009 05:44:13 +0000 (05:44 +0000)]
Eliminate an incorrect comment.

15 years agoDefine the kernel pmap in the same way on arm as on every other
Alan Cox [Thu, 7 May 2009 05:42:13 +0000 (05:42 +0000)]
Define the kernel pmap in the same way on arm as on every other
architecture.

Eliminate an unused definition.

Tested by: cognet

15 years agoRework r191742.
Dmitry Chagin [Thu, 7 May 2009 03:23:22 +0000 (03:23 +0000)]
Rework r191742.
Use the protocol family constants for the domain argument validation.

Return EAFNOSUPPORT in case when the incorrect domain argument
is specified.

Return EPROTONOSUPPORT instead of passing values that are not 0
to the BSD layer.

Suggested by:   rwatson

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

15 years agoAdd the Sierra AC885U and increase the max ports to 8.
Andrew Thompson [Thu, 7 May 2009 02:17:09 +0000 (02:17 +0000)]
Add the Sierra AC885U and increase the max ports to 8.

15 years ago- Fix the u3g port detection where it would not calculate the correct number of
Andrew Thompson [Thu, 7 May 2009 02:15:58 +0000 (02:15 +0000)]
- Fix the u3g port detection where it would not calculate the correct number of
  ports when multiple interfaces are present.
- Claim all interfaces regardless of how many are attached

15 years agoUse vendor and product macro expansion to make the device table smaller and
Andrew Thompson [Thu, 7 May 2009 02:13:56 +0000 (02:13 +0000)]
Use vendor and product macro expansion to make the device table smaller and
more readable.

15 years agooptimize ath_tx_findrix: there's no need to walk the rates table as
Sam Leffler [Thu, 7 May 2009 00:35:32 +0000 (00:35 +0000)]
optimize ath_tx_findrix: there's no need to walk the rates table as
sc_rixmap is an inverse map

NB: could eliminate the check for an invalid rate by filling in 0 for
    invalid entries but the rate control modules use it to identify
    bogus rates so leave it for now

15 years agoo cleanup checks for which vap combinations are permitted and what to
Sam Leffler [Wed, 6 May 2009 23:49:55 +0000 (23:49 +0000)]
o cleanup checks for which vap combinations are permitted and what to
  use for ic_opmode
o fixes the case where creating ahdemo+wds vaps caused ic_opmode to be
  set to hostap

15 years agoadd support for the Beacon Not Ready (BNR) interrupt
Sam Leffler [Wed, 6 May 2009 23:09:26 +0000 (23:09 +0000)]
add support for the Beacon Not Ready (BNR) interrupt
(available on 5211 and later)

15 years agoUse the good hints for the NSLU, it should fix the network adapter.
Olivier Houchard [Wed, 6 May 2009 20:24:17 +0000 (20:24 +0000)]
Use the good hints for the NSLU, it should fix the network adapter.

PR: arm/134092
Submitted by: gavin

15 years agofix grammar...
John-Mark Gurney [Wed, 6 May 2009 20:07:28 +0000 (20:07 +0000)]
fix grammar...

Reviewed by: cperciva

15 years ago- Split up the BIO queue into a queue for new and one for completed requests.
Ulf Lilleengen [Wed, 6 May 2009 19:34:32 +0000 (19:34 +0000)]
- Split up the BIO queue into a queue for new and one for completed requests.
  This is necessary for two reasons:
  1) In order to avoid collisions with the use of a BIOs flags set by a consumer
     or a provider
  2) Because GV_BIO_DONE was used to mark a BIO as done, not enough flags was
     available, so the consumer flags of a BIO had to be misused in order to
     support enough flags. The new queue makes it possible to recycle the
     GV_BIO_DONE flag into GV_BIO_GROW.
  As a consequence, gvinum will now work with any other GEOM class under it or
  on top of it.

- Use bio_pflags for storing internal flags on downgoing BIOs, as the requests
  appear to come from a consumer of a gvinum volume. Use bio_cflags only for
  cloned BIOs.
- Move gv_post_bio to be used internally for maintenance requests.
- Remove some cases where flags where set without need.

PR: kern/133604

15 years ago- Fix a case where a RAID5 volume would think that it is supposed to grow a new
Ulf Lilleengen [Wed, 6 May 2009 19:18:19 +0000 (19:18 +0000)]
- Fix a case where a RAID5 volume would think that it is supposed to grow a new
  subdisk after a parity rebuild.

15 years ago- Check if any plexes are doing internal maintenance before removing them.
Ulf Lilleengen [Wed, 6 May 2009 19:06:28 +0000 (19:06 +0000)]
- Check if any plexes are doing internal maintenance before removing them.

15 years ago- Add forgotten KASSERT.
Ulf Lilleengen [Wed, 6 May 2009 18:37:32 +0000 (18:37 +0000)]
- Add forgotten KASSERT.

15 years ago- Fix a bug where the bio_data field of the wrong BIO is freed if an error
Ulf Lilleengen [Wed, 6 May 2009 18:27:28 +0000 (18:27 +0000)]
- Fix a bug where the bio_data field of the wrong BIO is freed if an error
  occurs when doing a RAID5 request.

15 years ago- GV_BIO_RETRY is not used, and it is actually impossible with more than 8
Ulf Lilleengen [Wed, 6 May 2009 18:24:56 +0000 (18:24 +0000)]
- GV_BIO_RETRY is not used, and it is actually impossible with more than 8
  values for bio_cflags/bio_pflags.

15 years ago- Split the queue mutex into one for the event queue and one for the BIO queue,
Ulf Lilleengen [Wed, 6 May 2009 18:21:48 +0000 (18:21 +0000)]
- Split the queue mutex into one for the event queue and one for the BIO queue,
  as they do not really relate and to prepare for an additional queue to be
  covered by the BIO queue mutex.
- Implement wrappers for fetching the next element from the event queue as well
  as for putting a new element into the BIO queue.

15 years agoDisable adaptive mutexes and rwlocks for XENHVM.
Doug Rabson [Wed, 6 May 2009 17:52:38 +0000 (17:52 +0000)]
Disable adaptive mutexes and rwlocks for XENHVM.

15 years agoFix XENHVM build.
Doug Rabson [Wed, 6 May 2009 17:48:39 +0000 (17:48 +0000)]
Fix XENHVM build.

15 years agoRemove unnecessary CURVNET_SET() calls where curvnet context is
Marko Zec [Wed, 6 May 2009 13:30:46 +0000 (13:30 +0000)]
Remove unnecessary CURVNET_SET() calls where curvnet context is
(i.e. seems to be) already set.

This should reduce console noise due to curvnet recursion reports.

This change has no impact on nooptions VIMAGE builds.
Approved by: julian (mentor)

15 years agoUnbreak options VIMAGE kernel builds.
Marko Zec [Wed, 6 May 2009 08:49:39 +0000 (08:49 +0000)]
Unbreak options VIMAGE kernel builds.

Approved by: julian (mentor)

15 years agoReduce diffs against i386.
Yoshihiro Takahashi [Wed, 6 May 2009 04:40:18 +0000 (04:40 +0000)]
Reduce diffs against i386.
Use the hardclockintr function.

15 years agorevert r191633; this breaks at91 & xscale (likely all arm)
Sam Leffler [Wed, 6 May 2009 01:50:04 +0000 (01:50 +0000)]
revert r191633; this breaks at91 & xscale (likely all arm)

15 years agoo FreeBSD 7.2 added.
Maxim Konovalov [Tue, 5 May 2009 20:54:26 +0000 (20:54 +0000)]
o FreeBSD 7.2 added.

15 years agoAdd my birthday to the calendar.
Rene Ladan [Tue, 5 May 2009 16:45:25 +0000 (16:45 +0000)]
Add my birthday to the calendar.

Approved by: remko

15 years agoGPT style partitioning for loader(8) with U-Boot support library (tested on
Rafal Jaworowski [Tue, 5 May 2009 16:29:08 +0000 (16:29 +0000)]
GPT style partitioning for loader(8) with U-Boot support library (tested on
ARM).

Submitted by: Piotr Ziecik kosmo ! semihalf dot com

15 years agoSilence unsolicited spam printed out when KTR_MLD happens to be
Alexander Kabaev [Tue, 5 May 2009 16:27:45 +0000 (16:27 +0000)]
Silence unsolicited spam printed out when KTR_MLD happens to be
in KTR_COMPILE mask. Compiling KTR trace points in does not necessarily
mean enabling them, use proper check against ktr_mask instead.

15 years agoUnbreak LINT build, caused by a change in struct ng_node layout introduced
Marko Zec [Tue, 5 May 2009 16:26:06 +0000 (16:26 +0000)]
Unbreak LINT build, caused by a change in struct ng_node layout introduced
with r191816, which become uncovered only with NETGRAPH_DEBUG defined.

NOT approved by mentor (julian) due to emergency.

15 years agoMake sure the frame list base address is re-programmed after stopping the USB
Andrew Thompson [Tue, 5 May 2009 15:41:30 +0000 (15:41 +0000)]
Make sure the frame list base address is re-programmed after stopping the USB
schedule, in case the hardware clears the frame list base address.

Submitted by: Hans Petter Selasky
Reported by: Chao Shin

15 years agoRemove USB shutdown methods from device drivers as its the host controllers
Andrew Thompson [Tue, 5 May 2009 15:39:29 +0000 (15:39 +0000)]
Remove USB shutdown methods from device drivers as its the host controllers
responsibility to detach the bus.

PR: usb/133896
Submitted by: Hans Petter Selasky

15 years agoRevert part of r191494 which used the udev state to mark suspending, this needs
Andrew Thompson [Tue, 5 May 2009 15:36:23 +0000 (15:36 +0000)]
Revert part of r191494 which used the udev state to mark suspending, this needs
to be set via two variables (peer_suspended and self_suspended) and can not be
merged into one.

Submitted by: Hans Petter Selasky
Pointy hat: me

15 years agoAdd myself as a mentee of dwmalone for the committers graph. Also add my
Sean Nicholas Barkas [Tue, 5 May 2009 15:03:19 +0000 (15:03 +0000)]
Add myself as a mentee of dwmalone for the committers graph. Also add my
birthday to the calendar.

Approved by: dwmalone (mentor)

15 years agoMake gmultipath(8) easier to find.
Edward Tomasz Napierala [Tue, 5 May 2009 13:48:53 +0000 (13:48 +0000)]
Make gmultipath(8) easier to find.

15 years ago- Add support for PXA270 cpu.
Stanislav Sedov [Tue, 5 May 2009 12:57:16 +0000 (12:57 +0000)]
- Add support for PXA270 cpu.

Submitted by: Jacques Fourie <jacques.fourie@gmail.com>

15 years agoChange the curvnet variable from a global const struct vnet *,
Marko Zec [Tue, 5 May 2009 10:56:12 +0000 (10:56 +0000)]
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one.  The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE().  Recursions
on curvnet are permitted, though strongly discuouraged.

This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.

The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc.  Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.

The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.

This change also introduces a DDB subcommand to show the list of all
vnet instances.

Approved by: julian (mentor)

15 years agoAdd a constant PR_MAXMETHOD to better define the jail/OSD interface.
Jamie Gritton [Tue, 5 May 2009 05:49:08 +0000 (05:49 +0000)]
Add a constant PR_MAXMETHOD to better define the jail/OSD interface.

Reviewed by: dchagin, kib
Approved by: bz (mentor)

15 years agoDo not try to initialize LAPIC timer if we are not going to use it.
Alexander Motin [Tue, 5 May 2009 01:13:20 +0000 (01:13 +0000)]
Do not try to initialize LAPIC timer if we are not going to use it.
It solves assertion, when kernel built with INVARIANTS configured
to use i8254 timer.

15 years agoAlways compute the root of the kernel source tree and explicitly pass it
John Baldwin [Mon, 4 May 2009 20:25:56 +0000 (20:25 +0000)]
Always compute the root of the kernel source tree and explicitly pass it
to module builds.  This avoids having to have the module builds walk up
the tree to find the kernel sources.  It also allows a kernel + module
build to succeed when a new level of module subdirectories is added without
requiring that the /usr/share/mk/bsd.kmod.mk file on the machine be patched.

MFC after: 1 week

15 years agoMark Linux MIB sysctls MPSAFE.
Jamie Gritton [Mon, 4 May 2009 19:06:05 +0000 (19:06 +0000)]
Mark Linux MIB sysctls MPSAFE.

Reviewed by: dchagin, kib
Approved by: bz (mentor)

15 years agoOur grantpt(3) and unlockpt(3) don't comply with POSIX.
Ed Schouten [Mon, 4 May 2009 18:14:45 +0000 (18:14 +0000)]
Our grantpt(3) and unlockpt(3) don't comply with POSIX.

15 years agoUnlock the largest standard CPUID on Intel CPUs for both amd64 and i386 and
Jung-uk Kim [Mon, 4 May 2009 18:05:27 +0000 (18:05 +0000)]
Unlock the largest standard CPUID on Intel CPUs for both amd64 and i386 and
fix SMP topology detection.  On i386, we extend it to cover Core, Core 2,
and Core i7 processors, not just Pentium 4 family, and move it to better
place.  On amd64, all supported Intel CPUs should have this MSR.

15 years ago- Make the gvinum softc invisible to userland, as it is not needed.
Ulf Lilleengen [Mon, 4 May 2009 17:30:20 +0000 (17:30 +0000)]
- Make the gvinum softc invisible to userland, as it is not needed.

15 years ago Add the experimental nfs subtree to the kernel, that includes
Rick Macklem [Mon, 4 May 2009 15:23:58 +0000 (15:23 +0000)]
Add the experimental nfs subtree to the kernel, that includes
support for NFSv4 as well as NFSv2 and 3.
It lives in 3 subdirs under sys/fs:
nfs - functions that are common to the client and server
nfsclient - a mutation of sys/nfsclient that call generic functions
to do RPCs and handle state. As such, it retains the
buffer cache handling characteristics and vnode semantics that
are found in sys/nfsclient, for the most part.
nfsserver - the server. It includes a DRC designed specifically for
NFSv4, that is used instead of the generic DRC in sys/rpc.
The build glue will be checked in later, so at this point, it
consists of 3 new subdirs that should not affect kernel building.

Approved by: kib (mentor)

15 years agoRemove unneeded check for SESS_LEADER().
Ed Schouten [Mon, 4 May 2009 11:11:10 +0000 (11:11 +0000)]
Remove unneeded check for SESS_LEADER().

We perform the same check ~10 lines above.

15 years agoOops, sorry. Fix for fix.
Alexander Motin [Mon, 4 May 2009 08:41:54 +0000 (08:41 +0000)]
Oops, sorry. Fix for fix.

15 years agoThere is no atrtc driver in pc98, so hide atrtcclock_disable variable usage
Alexander Motin [Mon, 4 May 2009 08:36:47 +0000 (08:36 +0000)]
There is no atrtc driver in pc98, so hide atrtcclock_disable variable usage
in APM driver for this platform. This should fix pc98 build.

15 years agoFix whitespace and sorting in Symbol.map.
Ed Schouten [Mon, 4 May 2009 08:06:52 +0000 (08:06 +0000)]
Fix whitespace and sorting in Symbol.map.

15 years agoEliminate vnode_pager_input_smlfs()'s pointless call to pmap_clear_modify().
Alan Cox [Mon, 4 May 2009 06:30:00 +0000 (06:30 +0000)]
Eliminate vnode_pager_input_smlfs()'s pointless call to pmap_clear_modify().
The page can't possibly have any modified page table entries because it
isn't even mapped.

15 years agoRemove redundant NFSMNT_NFSV3 check in DTrace hooks for NFS RPC.
Robert Watson [Mon, 4 May 2009 02:19:52 +0000 (02:19 +0000)]
Remove redundant NFSMNT_NFSV3 check in DTrace hooks for NFS RPC.

MFC after: 1 month

15 years agoFix typo in comment.
Robert Watson [Mon, 4 May 2009 02:06:39 +0000 (02:06 +0000)]
Fix typo in comment.

MFC after: 1 month

15 years agoThis belongs in //svn.freebsd.org/base/vendor/file/dist now.
David E. O'Brien [Mon, 4 May 2009 00:42:15 +0000 (00:42 +0000)]
This belongs in //svn.freebsd.org/base/vendor/file/dist now.

15 years agoMerge vendor/file/dist@191739, bringing FILE 5.00 to 8-CURRENT.
David E. O'Brien [Mon, 4 May 2009 00:37:44 +0000 (00:37 +0000)]
Merge vendor/file/dist@191739, bringing FILE 5.00 to 8-CURRENT.

15 years agoRelax the condition for printing the lost state transition message. The new
Andrew Thompson [Sun, 3 May 2009 18:29:04 +0000 (18:29 +0000)]
Relax the condition for printing the lost state transition message. The new
state will be set before the EXT_STATEWAIT flag is cleared and its ok to
transition again at that point.