]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
14 years agoUpdate to a 26-Nov-2009 release.
Ruslan Ermilov [Sat, 9 Jan 2010 23:19:01 +0000 (23:19 +0000)]
Update to a 26-Nov-2009 release.

14 years agoVendor import of bwk's 26-Nov-2009 release.
Ruslan Ermilov [Sat, 9 Jan 2010 23:04:24 +0000 (23:04 +0000)]
Vendor import of bwk's 26-Nov-2009 release.

14 years agoRemove file system support based on the simple file system protocol
Marcel Moolenaar [Sat, 9 Jan 2010 22:54:29 +0000 (22:54 +0000)]
Remove file system support based on the simple file system protocol
as this only allows us to access file systems that EFI knows about.
With a loader that can only use EFI-supported file systems, we're
forced to put /boot on the EFI system partition. This is suboptimal
in the following ways:
1.  With /boot a symlink to /efi/boot, mergemaster complains about
    the mismatch and there's no quick solution.
2.  The EFI loader can only boot a single version of FreeBSD. There's
    no way to install multiple versions of FreeBSD and select one
    at the loader prompt.
3.  ZFS maintains /boot/zfs/zpool.cache and with /boot a symlink we
    end up with the file on a MSDOS file system. ZFS does not have
    proper handling of file systems that are under Giant.

Implement a disk device based on the block I/O protocol instead and
pull in file system code from libstand. The disk devices are really
the partitions that EFI knows about.

This change is backward compatible.

MFC after: 1 week

14 years agoClean up import.
Ruslan Ermilov [Sat, 9 Jan 2010 22:47:40 +0000 (22:47 +0000)]
Clean up import.

14 years agoClean up.
Ruslan Ermilov [Sat, 9 Jan 2010 22:39:01 +0000 (22:39 +0000)]
Clean up.

14 years agoImplement the fo_readdir method. This does not support long
Marcel Moolenaar [Sat, 9 Jan 2010 22:33:34 +0000 (22:33 +0000)]
Implement the fo_readdir method. This does not support long
file names.

Obtained from: Juniper Networks, Inc.
MFC after: 1 week

14 years agoFlatten out vendor tree.
Ruslan Ermilov [Sat, 9 Jan 2010 22:31:11 +0000 (22:31 +0000)]
Flatten out vendor tree.

14 years agoLong ago, in r120654, the rounding of KERNend and physfree in locore
Alan Cox [Sat, 9 Jan 2010 22:09:10 +0000 (22:09 +0000)]
Long ago, in r120654, the rounding of KERNend and physfree in locore
was changed from a small page boundary to a large page boundary.  As
a consequence pmap_kmem_choose() became a pointless waste of address
space.  Eliminate it.

14 years ago- Add code allowing a network device to only be open and closed once
Marius Strobl [Sat, 9 Jan 2010 21:23:39 +0000 (21:23 +0000)]
- Add code allowing a network device to only be open and closed once
  by keeping it opened after the first open and closing it via the
  cleanup handler when NETIF_OPEN_CLOSE_ONCE is defined in order to
  avoid the open-close-dance on every file access which with firmware
  that for example performs an auto-negotiation on every open causes
  netbooting to take horribly long. Basically the behavior with this
  knob enabled resembles the one employed between r60506 and r177108
  (and for sparc64 also again since r182919) with the addition that
  the network device now is closed eventually before entering the
  kernel and before rebooting. Actually I think this should be the
  desired MI behavior, however the U-Boot loader actually requires
  net_close() to be called after every transaction in order for some
  local shutdown operations to be performed (and which I think thus
  will break on concurrent opens, i.e. when netdev_opens is > 1, like
  the loader does at least for disks when LOADER_GZIP_SUPPORT is
  enabled).
- Use NETIF_OPEN_CLOSE_ONCE to replace the hack, which artificially
  increased netdev_opens for sparc64 in order to keep the network
  device opened forever, as at least some firmware versions require
  the network device to be closed eventually before entering the
  kernel or otherwise will DMA received packets to stale memory.
  The powerpc OFW loader probably wants NETIF_OPEN_CLOSE_ONCE to be
  set as well for the same reasons.

14 years agoUpdate files to remove when MK_BLUETOOTH=no.
Antoine Brodin [Sat, 9 Jan 2010 19:52:42 +0000 (19:52 +0000)]
Update files to remove when MK_BLUETOOTH=no.

14 years agoThe client type rule allows DHCP, implicitly. Since DHCPv6 uses
Hajimu UMEMOTO [Sat, 9 Jan 2010 19:16:27 +0000 (19:16 +0000)]
The client type rule allows DHCP, implicitly.  Since DHCPv6 uses
link-local address unlike with DHCP, we need one more rule to allow
the DHCPv6.

Reported by: David Horn <dhorn2000__at__gmail.com>

14 years agoAdd files to remove when MK_AUTHPF=no.
Antoine Brodin [Sat, 9 Jan 2010 19:13:56 +0000 (19:13 +0000)]
Add files to remove when MK_AUTHPF=no.

14 years agoAdd files to remove when MK_ATM=no.
Antoine Brodin [Sat, 9 Jan 2010 19:12:33 +0000 (19:12 +0000)]
Add files to remove when MK_ATM=no.

14 years agoAdd files to remove when MK_AT=no.
Antoine Brodin [Sat, 9 Jan 2010 19:09:16 +0000 (19:09 +0000)]
Add files to remove when MK_AT=no.

14 years agoAdd files to remove when MK_APM=no.
Antoine Brodin [Sat, 9 Jan 2010 19:07:04 +0000 (19:07 +0000)]
Add files to remove when MK_APM=no.

14 years agoAdd files to remove when MK_AMD=no.
Antoine Brodin [Sat, 9 Jan 2010 19:04:38 +0000 (19:04 +0000)]
Add files to remove when MK_AMD=no.

14 years agoSend link state change control messages to "orphans" hook as well.
Max Khon [Sat, 9 Jan 2010 19:03:48 +0000 (19:03 +0000)]
Send link state change control messages to "orphans" hook as well.

MFC after: 1 week

14 years agoAdd files to remove when MK_ACCT=no.
Antoine Brodin [Sat, 9 Jan 2010 19:02:33 +0000 (19:02 +0000)]
Add files to remove when MK_ACCT=no.

14 years agoAdd missing library dependency.
Antoine Brodin [Sat, 9 Jan 2010 18:59:03 +0000 (18:59 +0000)]
Add missing library dependency.

MFC after: 1 month

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:59:03 +0000 (18:59 +0000)]
Merge from projects/mips to head by hand:

Copy the support files for the Octeon 1 CPU from sys/mips/octeon1 on
the projects/mips side to sys/mips/cavium on the head side to conform
to the other vendor code.  This code was contributed by Cavium to the
project and forward ported by Warner Losh, with some additional code
from Randal Stewart.

# I'll fix the building problems the move creates in a future commit.

14 years agolibusb20 was renamed libusb several months ago.
Antoine Brodin [Sat, 9 Jan 2010 18:55:29 +0000 (18:55 +0000)]
libusb20 was renamed libusb several months ago.

MFC after: 1 month

14 years agoFix a typo.
Antoine Brodin [Sat, 9 Jan 2010 18:53:03 +0000 (18:53 +0000)]
Fix a typo.

MFC after: 1 month

14 years agoFix a typo.
Antoine Brodin [Sat, 9 Jan 2010 18:51:50 +0000 (18:51 +0000)]
Fix a typo.

MFC after: 1 month

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:29:35 +0000 (18:29 +0000)]
Merge from projects/mips to head by hand:

Copy over the support files (except sys/conf and sys/mips/conf) for
RMI XLR processor support.  This port has been contributed by RMI and
brought up to date by Randal Stewart (rrs@).  This port is a work in
progress, and there might still be significant changes.  The port
makes it to multi-user, but is still early beta.

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:19:31 +0000 (18:19 +0000)]
Merge from projects/mips to head by hand:

Copy over MALTA64, the 64-bit varianat of the malta board...

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:17:39 +0000 (18:17 +0000)]
Merge from projects/mips to head by hand:

Copy over XLR kernel config file.

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:15:28 +0000 (18:15 +0000)]
Merge from projects/mips to head by hand:

Copy over the OCTEON1 kernel config file.   This is the 64-bit version.

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:14:27 +0000 (18:14 +0000)]
Merge from projects/mips to head by hand:

Copy over OCTEON1.hints file.

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:13:13 +0000 (18:13 +0000)]
Merge from projects/mips to head by hand:

Copy over OCTEON1-32 file: the 32-bit variant of the octeon kernel
config file.

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:11:45 +0000 (18:11 +0000)]
Merge from projects/mips to head by hand:

Copy over AR71XX.hints file.

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:10:46 +0000 (18:10 +0000)]
Merge from projects/mips to head by hand:

Copy over the AR71XX config file.

14 years agoMerge from projects/mips to head by hand.
Warner Losh [Sat, 9 Jan 2010 18:09:30 +0000 (18:09 +0000)]
Merge from projects/mips to head by hand.

Copy over the SWARM.hints file.

14 years agoMerge from pprojects/mips to head by hand.
Warner Losh [Sat, 9 Jan 2010 18:08:31 +0000 (18:08 +0000)]
Merge from pprojects/mips to head by hand.

Copy over the SWARM config file.

14 years agoExtra parantheses to keep certain compilers happy.
Navdeep Parhar [Sat, 9 Jan 2010 18:07:10 +0000 (18:07 +0000)]
Extra parantheses to keep certain compilers happy.

Submitted by: trasz@

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 18:02:31 +0000 (18:02 +0000)]
Merge from projects/mips to head by hand:

Merge support files for the Atheros AR71xx (and soon AR9xxx)
processors, except files from sys/conf and sys/mips/conf.  This work
was done primarily by Olecksandr Tymoshenko and works on the
RouterStation and RouterStation PRO.  Other AR71xx-based boards have
been reported as working as well (RouterBoard, for example).

14 years agoMerge from projects/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 17:56:25 +0000 (17:56 +0000)]
Merge from projects/mips to head by hand:

Copy the files for the sibyte support (except files in sys/conf and
sys/mips/conf).  This targets the Broadcom SWARM board (bcm91250) and
the SB-1 core in the BCM1250 SoC.  This work was done by Neel Natu.

14 years agoMerge r201902 and r195669 from projects/mips into head by hand:
Warner Losh [Sat, 9 Jan 2010 17:21:36 +0000 (17:21 +0000)]
Merge r201902 and r195669 from projects/mips into head by hand:

r201902 | imp | 2010-01-09 10:16:19 -0700 (Sat, 09 Jan 2010) | 2 lines
Fix comment, which was missed in an earlier commit...

r195669 | gonzo | 2009-07-13 17:03:44 -0600 (Mon, 13 Jul 2009) | 3 lines
- Remove -mno-dsp from CFLAGS. MIPS DSP ASE is off by default
   now (as it should be)

14 years agoFix comment, which was missed in an earlier commit...
Warner Losh [Sat, 9 Jan 2010 17:16:19 +0000 (17:16 +0000)]
Fix comment, which was missed in an earlier commit...

14 years agoRemove clause 3 and 4 from TNF licenses (this was the only 4-clause TNF
Marius Strobl [Sat, 9 Jan 2010 17:09:57 +0000 (17:09 +0000)]
Remove clause 3 and 4 from TNF licenses (this was the only 4-clause TNF
license FreeBSD had in sys/boot).

Obtained from: NetBSD

14 years agoSome style(9) fixes in order to fabricate a commit to denote that
Marius Strobl [Sat, 9 Jan 2010 15:59:15 +0000 (15:59 +0000)]
Some style(9) fixes in order to fabricate a commit to denote that
the commit message for r201896 actually should have read:

As nfsm_srvmtofh_xx() assumes the 4-byte alignment required by XDR
ensure the mbuf data is aligned accordingly by calling nfs_realign()
in fha_extract_info(). This fix is orthogonal to the problem solved
by r199274/r199284.

PR: 142102 (second part)
MFC after: 1 week

14 years agoAdd comments trying to explain what bad things happen here, i.e.
Bjoern A. Zeeb [Sat, 9 Jan 2010 15:43:47 +0000 (15:43 +0000)]
Add comments trying to explain what bad things happen here, i.e.
how hashed MD5/SHA are implemented, abusing Final() for padding and
sw_octx to transport the key from the beginning to the end.

Enlightened about what was going on here by: cperciva
Reviewed by: cperciva
MFC After: 3 days
X-MFC with: r187826
PR: kern/126468

14 years agoCorrect spelling.
Bjoern A. Zeeb [Sat, 9 Jan 2010 15:37:24 +0000 (15:37 +0000)]
Correct spelling.

Submitted by: (pluknet gmail.com)
MFC after: 4 days
X-MFC with: r201806

14 years agoExclude options COMPAT_FREEBSD4 now that the MD freebsd4_sigreturn()
Marius Strobl [Sat, 9 Jan 2010 15:31:27 +0000 (15:31 +0000)]
Exclude options COMPAT_FREEBSD4 now that the MD freebsd4_sigreturn()
is gone since r201396 and which is also in line with the fact that
FreeBSD 4 didn't supported sparc64.

PR: 142102 (second part)
MFC after: 1 week

14 years agoReduce recursions on curvnet and thus spamming the console with warning
Marko Zec [Sat, 9 Jan 2010 14:56:38 +0000 (14:56 +0000)]
Reduce recursions on curvnet and thus spamming the console with warning
messages for kernels built with options VIMAGE and VNET_DEBUG enabled.

Reviewed by: bz
MFC after: 3 days

14 years agoFix formatting.
Xin LI [Sat, 9 Jan 2010 12:34:15 +0000 (12:34 +0000)]
Fix formatting.

MFC after: 2 weeks

14 years agoAdd a set of manual pages for pthread[_attr]_[sg]etaffinity(3).
Xin LI [Sat, 9 Jan 2010 12:31:11 +0000 (12:31 +0000)]
Add a set of manual pages for pthread[_attr]_[sg]etaffinity(3).

Reviewed by: davidxu
MFC after: 2 weeks

14 years agoSet md_ldt (pointer to the LDT) after md_ldt_sd (system segment
Konstantin Belousov [Sat, 9 Jan 2010 11:28:01 +0000 (11:28 +0000)]
Set md_ldt (pointer to the LDT) after md_ldt_sd (system segment
descriptor for the LDT) is populated. md_ldt is used by context-switch
code as indicator that LDT segment register shall be loaded with
GUSERLDT segment instead of 0, so context switch at the wrong time may
cause attempt to load non-populated descriptor.

Use store with the barrier to prevent other CPUs from seeing updated
md_ldt but not seeing updated md_ldt_sd. Multithreaded process may
context-switch to another thread of the process on another CPU and read
md_ldt.

MFC after: 1 week

14 years agoVarious language fixes. Also fixed the URL to totd, obtained from NetBSD.
Christian Brueffer [Sat, 9 Jan 2010 10:24:09 +0000 (10:24 +0000)]
Various language fixes.  Also fixed the URL to totd, obtained from NetBSD.

PR: 38061
Submitted by: Chris Pepper <pepper@rockefeller.edu>
MFC after: 1 week

14 years agobridge(4) acts like a switch, not like a hub.
Christian Brueffer [Sat, 9 Jan 2010 09:30:09 +0000 (09:30 +0000)]
bridge(4) acts like a switch, not like a hub.

PR: 141350
Submitted by: brucec
Reviewed by: thompsa
MFC after: 1 week

14 years agoUse enum to define key types.
David Xu [Sat, 9 Jan 2010 06:30:40 +0000 (06:30 +0000)]
Use enum to define key types.

Suggested by: jmallett

14 years agoput semaphore waiter in long term list.
David Xu [Sat, 9 Jan 2010 06:12:44 +0000 (06:12 +0000)]
put semaphore waiter in long term list.

14 years agoAdd key type TYPE_SEM.
David Xu [Sat, 9 Jan 2010 06:05:31 +0000 (06:05 +0000)]
Add key type TYPE_SEM.

14 years agoUpdate manual for sem_init and sem_open, restrictions of shared semaphore
David Xu [Sat, 9 Jan 2010 05:40:46 +0000 (05:40 +0000)]
Update manual for sem_init and sem_open, restrictions of shared semaphore
and 14 characters in name length no longer exist.

Reviewed by: deischen@

14 years agoMerge r195128 from project/mips to head.
Warner Losh [Sat, 9 Jan 2010 04:59:57 +0000 (04:59 +0000)]
Merge r195128 from project/mips to head.

r195128 | gonzo | 2009-06-27 17:27:41 -0600 (Sat, 27 Jun 2009) | 4 lines
- Add support for handling TLS area address in kernel space.
    From the userland point of view get/set operations are
    performed using sysarch(2) call.

14 years agoRevert iwn channel-argument handling of set_txpower() to r201822
Giorgos Keramidas [Sat, 9 Jan 2010 04:00:22 +0000 (04:00 +0000)]
Revert iwn channel-argument handling of set_txpower() to r201822

Pass the channel argument as a real argument and not through
RXON.  The RXON version seems to have problem with both the older
228.57.2.23 firmware and the latest 228.61.2.24 version resulting
in device initialization errors like:

  wpa_supplicant[2928]: Failed to initiate AP scan.
  kernel: firmware error log:
  kernel: error type      = "SYSASSERT" (0x00000005)
  kernel: program counter = 0x0000147C
  kernel: source line     = 0x0000058B
  kernel: error data      = 0x0000058B00000000
  kernel: branch link     = 0x0000145A00001492
  kernel: interrupt link  = 0x000006DE00000000
  kernel: time            = 7310
  ...
  kernel: iwn0: iwn_config: could not set TX power
  kernel: iwn0: iwn_init_locked: could not configure device, error 35

By passing the current channel to hal->set_txpower() the firmware
error is fixed, at least for the 4965 chipset of my Thinkpad and
the ones tested by Bernhard.

Submitted by: Bernhard Schmidt <bschmidt@techwires.net>
Approved by: rpaulo

14 years agoRename mips_pcpu_init to mips_pcpu0_init since it applies only to the
Warner Losh [Sat, 9 Jan 2010 03:08:22 +0000 (03:08 +0000)]
Rename mips_pcpu_init to mips_pcpu0_init since it applies only to the
BSP.  Provide a missing prototype.

14 years agoCompute the target of the jump in the 'J' and 'JAL' instructions
Neel Natu [Sat, 9 Jan 2010 02:17:14 +0000 (02:17 +0000)]
Compute the target of the jump in the 'J' and 'JAL' instructions
correctly. The 256MB segment is formed by taking the top 4 bits
of the address of the instruction in the "branch delay" slot
as opposed to the 'J' or 'JAL' instruction itself.

Approved by: imp (mentor)

14 years agoIntroduce the new kernel thread called "deadlock resolver".
Attilio Rao [Sat, 9 Jan 2010 01:46:38 +0000 (01:46 +0000)]
Introduce the new kernel thread called "deadlock resolver".
While the name is pretentious, a good explanation of its targets is
reported in this 17 months old presentation e-mail:
http://lists.freebsd.org/pipermail/freebsd-arch/2008-August/008452.html

In order to implement it, the sq_type in sleepqueues is mandatory and not
only compiled along with INVARIANTS option. Additively, a new sleepqueue
function, sleepq_type() is added, returning the type of the sleepqueue
linked to a wchan.
Three new sysctls are added in order to configure the thread:
debug.deadlkres.slptime_threshold
debug.deadlkres.blktime_threshold
debug.deadlkres.sleepfreq

rappresenting the thresholds for sleep and block time that will lead to
a deadlock matching (when exceeded), while the sleepfreq rappresents the
number of seconds between 2 consecutive thread runnings.
In order to enable the deadlock resolver thread recompile your kernel
with the option DEADLKRES.

Reviewed by: jeff
Tested by: pho, Giovanni Trematerra
Sponsored by: Nokia Incorporated, Sandvine Incorporated
MFC after: 2 weeks

14 years agoVarious language fixes.
Christian Brueffer [Sat, 9 Jan 2010 00:24:54 +0000 (00:24 +0000)]
Various language fixes.

PR: 38061
Submitted by: Chris Pepper <pepper@rockefeller.edu>
MFC after: 1 week

14 years agoMerge r195129 from project/mips to head by hand:
Warner Losh [Sat, 9 Jan 2010 00:07:47 +0000 (00:07 +0000)]
Merge r195129 from project/mips to head by hand:

r195129 | gonzo | 2009-06-27 17:28:56 -0600 (Sat, 27 Jun 2009) | 2 lines
- Use sysarch(2) in MIPS version of _tcb_set/_tcb_get

14 years agoMerge r195030 from project/mips to head by hand
Warner Losh [Sat, 9 Jan 2010 00:01:35 +0000 (00:01 +0000)]
Merge r195030 from project/mips to head by hand

r195030 | gonzo | 2009-06-25 19:27:31 -0600 (Thu, 25 Jun 2009) | 4 lines
- Switch to libc softfloat from libgcc implementation. The problem
    with latter is that it is not complete, fpsetXXX/fpgetXXX
    functions are missing.

14 years agoMerge r197800 from projects/mips to head by hand:
Warner Losh [Fri, 8 Jan 2010 23:59:04 +0000 (23:59 +0000)]
Merge r197800 from projects/mips to head by hand:

r197800 | gonzo | 2009-10-06 00:35:52 -0600 (Tue, 06 Oct 2009) | 3 lines
- curbrk variable for sbrk and brk should be the same
- Add correct variable names to Symbol.map

14 years agoMerge r195025 from projects/mips to head by hand:
Warner Losh [Fri, 8 Jan 2010 23:50:39 +0000 (23:50 +0000)]
Merge r195025 from projects/mips to head by hand:

r195025 | gonzo | 2009-06-25 19:01:50 -0600 (Thu, 25 Jun 2009) | 4 lines
- Move fpgetXXX.c/fpsetXXX.c sources to hardfloat subdir/
    to prevenmt them from being mixed up with lib/libc/softfloat
    files with the same names

14 years agoMerge r194519 from projects/mips to head by hand:
Warner Losh [Fri, 8 Jan 2010 23:37:30 +0000 (23:37 +0000)]
Merge r194519 from projects/mips to head by hand:

r194519 | gonzo | 2009-06-19 17:28:26 -0600 (Fri, 19 Jun 2009) | 3 lines
- set -mabicalls and -msoft-float as a default in order to
    simplify building ports

14 years agoMake options KGSSAPI build and add it to NOTES.
Brooks Davis [Fri, 8 Jan 2010 23:26:10 +0000 (23:26 +0000)]
Make options KGSSAPI build and add it to NOTES.

rpcsec_gss_prot.c:
  Use kernel printf and headers.

vc_rpcsec_gss.c:
  Use a local RPCAUTH_UNIXGIDS definition for 16 instead of using NGROUPS.

14 years agoMerge r195030 from project/mips into head by hand:
Warner Losh [Fri, 8 Jan 2010 23:11:23 +0000 (23:11 +0000)]
Merge r195030 from project/mips into head by hand:

r195030 | gonzo | 2009-06-25 19:27:31 -0600 (Thu, 25 Jun 2009) | 4 lines
- Switch to libc softfloat from libgcc implementation. The problem
    with latter is that it is not complete, fpsetXXX/fpgetXXX
    functions are missing.

14 years agoMerge r195575 and 195530 from projects/mips to head by hand:
Warner Losh [Fri, 8 Jan 2010 23:09:11 +0000 (23:09 +0000)]
Merge r195575 and 195530 from projects/mips to head by hand:

r195575 | imp | 2009-07-10 12:24:02 -0600 (Fri, 10 Jul 2009) | 2 lines
quick hack for the problem gonzo is seeing.

r195530 | imp | 2009-07-10 01:18:30 -0600 (Fri, 10 Jul 2009) | 5 lines
Always build all 4 emulators into the mips toolchain.
# I think we have a gcc spec file issue with abi=64 since I have to do other
# hacks to get it mostly kinda right.

14 years agoMerge from project/mips r195021:
Warner Losh [Fri, 8 Jan 2010 23:06:56 +0000 (23:06 +0000)]
Merge from project/mips r195021:

r195021 | gonzo | 2009-06-25 18:44:23 -0600 (Thu, 25 Jun 2009) | 2 lines
- Add MIPS to the list of 32-bit architectures

14 years agoBack-merge (by hand) r195668:
Warner Losh [Fri, 8 Jan 2010 23:02:17 +0000 (23:02 +0000)]
Back-merge (by hand) r195668:

r195668 | gonzo | 2009-07-13 17:01:12 -0600 (Mon, 13 Jul 2009) | 3 lines
- Get rid of ugly TARGET_CPU_DEFAULT default. 16 is MASK_DSP
    and was set there due to my ignroance.

14 years agoFree allocated sbufs before returning ENOMEM.
Christian Brueffer [Fri, 8 Jan 2010 22:58:50 +0000 (22:58 +0000)]
Free allocated sbufs before returning ENOMEM.

PR: 128335
Submitted by: Mateusz Guzik <mjguzik@gmail.com>
MFC after: 2 week

14 years agoRevert r199758. It pointed out that we were calling pcpu_init way too
Warner Losh [Fri, 8 Jan 2010 22:52:02 +0000 (22:52 +0000)]
Revert r199758.  It pointed out that we were calling pcpu_init way too
late...

14 years agoCentralize initialization of pcpu, and set curthread early...
Warner Losh [Fri, 8 Jan 2010 22:48:21 +0000 (22:48 +0000)]
Centralize initialization of pcpu, and set curthread early...

14 years agoRemove unnecessary quoting and markup, add missing punctuation.
Christian Brueffer [Fri, 8 Jan 2010 22:02:42 +0000 (22:02 +0000)]
Remove unnecessary quoting and markup, add missing punctuation.

PR: 140494
Submitted by: Jeremy Huddleston <jeremyhu@apple.com>, bde
MFC after: 1 week

14 years agoFix array overflow.
Edward Tomasz Napierala [Fri, 8 Jan 2010 21:02:12 +0000 (21:02 +0000)]
Fix array overflow.

Reviewed by: rpaulo

14 years agoGet rid of #ident.
Edward Tomasz Napierala [Fri, 8 Jan 2010 20:40:28 +0000 (20:40 +0000)]
Get rid of #ident.

14 years agosk(4), stge(4) and vge(4) should work on all architectures, snd_t4dwave(4)
Christian Brueffer [Fri, 8 Jan 2010 20:38:41 +0000 (20:38 +0000)]
sk(4), stge(4) and vge(4) should work on all architectures, snd_t4dwave(4)
works on sparc64.

Submitted by: yongari

14 years agoFix array overflow.
Edward Tomasz Napierala [Fri, 8 Jan 2010 19:56:37 +0000 (19:56 +0000)]
Fix array overflow.

Reviewed by: philip

14 years agoTo avoid hardcoding further kernel configuration names for
Bjoern A. Zeeb [Fri, 8 Jan 2010 19:01:50 +0000 (19:01 +0000)]
To avoid hardcoding further kernel configuration names for
make universe, split the logic into two parts:
- 1st to build worlds and generate kernel configs like LINT.
- 2nd to build kernels for a given TARGET architecture correctly
  finding all newly generated configs, not knowing anything about
  LINT anymore. (*)

(*) If you know better/cleaner/... ways to do this, let me know.

Discussed on/with: arch, jhb, rwatson
MFC after: 1 month

14 years agoGenerate a second LINT configuration for i386 and amd64 in
Bjoern A. Zeeb [Fri, 8 Jan 2010 19:00:20 +0000 (19:00 +0000)]
Generate a second LINT configuration for i386 and amd64 in
sys/conf/makeLINT.mk, which includes LINT and sets options VIMAGE
so that we will have VIMAGE LINT builds[1]. For now only do it for
those two architectures to avoid massive universe times for archs,
where people will less likely use VIMAGE or not at all.

Requested by: jhb [1]
Discussed on/with: arch, jhb, rwatson
MFC after: 1 month

14 years agoIn sys/<arch>/conf/Makefile set TARGET to <arch>. That allows
Bjoern A. Zeeb [Fri, 8 Jan 2010 18:57:31 +0000 (18:57 +0000)]
In sys/<arch>/conf/Makefile set TARGET to <arch>. That allows
sys/conf/makeLINT.mk to only do certain things for certain
architectures.

Note that neither arm nor mips have the Makefile there, thus
essentially not (yet) supporting LINT.  This would enable them
do add special treatment to sys/conf/makeLINT.mk as well chosing
one of the many configurations as LINT.

This is a hack of doing this and keeping it in a separate commit
will allow us to more easily identify and back it out.

Discussed on/with: arch, jhb (as part of the LINT-VIMAGE thread)
MFC after: 1 month

14 years agoEnsure an address is removed from the interface address
Qing Li [Fri, 8 Jan 2010 17:49:24 +0000 (17:49 +0000)]
Ensure an address is removed from the interface address
list when the installation of that address fails.

PR: 139559

14 years agoRemove ste(4), it should work on all architectures now.
Christian Brueffer [Fri, 8 Jan 2010 17:47:37 +0000 (17:47 +0000)]
Remove ste(4), it should work on all architectures now.

14 years agoGet rid of #ident, GCC 4.4 warns about it being deprecated.
Edward Tomasz Napierala [Fri, 8 Jan 2010 17:20:41 +0000 (17:20 +0000)]
Get rid of #ident, GCC 4.4 warns about it being deprecated.

OK-ed by: scottl

14 years agoSwitch traceroute over to make use of proper in-kernel source
Bjoern A. Zeeb [Fri, 8 Jan 2010 16:59:28 +0000 (16:59 +0000)]
Switch traceroute over to make use of proper in-kernel source
address selection.

Reviewed by: rwatson, fenner
MFC after: 1 week
X-MFCable to: stable/8, stable/7 (after r183571).
PR: kern/139454
Tested by: Frank Steinborn (steinex nognu.de)

14 years agoUnbreak the XEN build after r201751.
Bjoern A. Zeeb [Fri, 8 Jan 2010 16:56:11 +0000 (16:56 +0000)]
Unbreak the XEN build after r201751.

14 years agoStop GCC from complaining about lagg_port_checkstacking() being unused.
Edward Tomasz Napierala [Fri, 8 Jan 2010 16:44:33 +0000 (16:44 +0000)]
Stop GCC from complaining about lagg_port_checkstacking() being unused.

14 years agoComplete the swap of carp(4) log levels and document the change.
Ruslan Ermilov [Fri, 8 Jan 2010 16:14:41 +0000 (16:14 +0000)]
Complete the swap of carp(4) log levels and document the change.

MFC after: 3 days

14 years agoOne more #ifdef fix for GCC 4.4.
Edward Tomasz Napierala [Fri, 8 Jan 2010 15:57:56 +0000 (15:57 +0000)]
One more #ifdef fix for GCC 4.4.

14 years agoRemove unused smbfs_smb_qpathinfo().
Edward Tomasz Napierala [Fri, 8 Jan 2010 15:53:07 +0000 (15:53 +0000)]
Remove unused smbfs_smb_qpathinfo().

14 years agoRemove unused uhci_dump_qhs().
Edward Tomasz Napierala [Fri, 8 Jan 2010 15:52:06 +0000 (15:52 +0000)]
Remove unused uhci_dump_qhs().

Reviewed by: hps

14 years agoRemove unused iir_watchdog().
Edward Tomasz Napierala [Fri, 8 Jan 2010 15:47:23 +0000 (15:47 +0000)]
Remove unused iir_watchdog().

Reviewed by: scottl

14 years agoPut mly_timeout() under MLY_DEBUG, so that newer GCC versions don't
Edward Tomasz Napierala [Fri, 8 Jan 2010 15:46:27 +0000 (15:46 +0000)]
Put mly_timeout() under MLY_DEBUG, so that newer GCC versions don't
complain about unused static function.

Reviewed by: scottl

14 years agoReplace several instances of 'if (!a & b)' with 'if (!(a &b))' in order
Edward Tomasz Napierala [Fri, 8 Jan 2010 15:44:49 +0000 (15:44 +0000)]
Replace several instances of 'if (!a & b)' with 'if (!(a &b))' in order
to silence newer GCC versions.

14 years agoFix #ifdefs so that GCC 4.4 doesn't complain about it.
Edward Tomasz Napierala [Fri, 8 Jan 2010 15:41:24 +0000 (15:41 +0000)]
Fix #ifdefs so that GCC 4.4 doesn't complain about it.

Reviewed by: rpaulo

14 years ago- Fix a bug in sched_4bsd where the timestamp for the sleeping operation
Attilio Rao [Fri, 8 Jan 2010 14:55:11 +0000 (14:55 +0000)]
- Fix a bug in sched_4bsd where the timestamp for the sleeping operation
  is not cleaned up on the wakeup but reset.
  This is harmless mostly because td_slptick (and ki_slptime from
  userland) should be analyzed only with the assumption that the thread
  is actually sleeping (thus while the td_slptick is correctly set) but
  without this invariant the number is nomore consistent.
- Move td_slptick from u_int to int in order to follow 'ticks' signedness
  and wrap up accordingly [0]

[0] Submitted by: emaste
Sponsored by: Sandvine Incorporated
MFC 1 week

14 years agoThis now passes a make universe with WARNS=6.
David Malone [Fri, 8 Jan 2010 11:16:58 +0000 (11:16 +0000)]
This now passes a make universe with WARNS=6.

14 years agoAllow comment (#) to be placed anywhere in the line, not only at the
Maxim Sobolev [Fri, 8 Jan 2010 10:54:15 +0000 (10:54 +0000)]
Allow comment (#) to be placed anywhere in the line, not only at the
beginning, so it's consistent with other configuration files.

MFC after: 3 weeks

14 years agoIf we fail to read the Ethernet address from the card, just print an
Gavin Atkinson [Fri, 8 Jan 2010 10:32:27 +0000 (10:32 +0000)]
If we fail to read the Ethernet address from the card, just print an
warning message and attach without setting the Ethernet address to a
random address.  It is not believed that this code can actually be
executed, and if it does, we're better off printing an error message than
faking up an Ethernet address.

PR: kern/133239
Reviewed by: yongari (earlier version of patch)
Approved by: ed (mentor)

14 years agoCatch up with r183101 that added "device acpi" to GENERIC.
Alan Cox [Fri, 8 Jan 2010 09:16:37 +0000 (09:16 +0000)]
Catch up with r183101 that added "device acpi" to GENERIC.

14 years ago- Change the type of size_max to u_quad_t because its value is converted
Jaakko Heinonen [Fri, 8 Jan 2010 07:57:43 +0000 (07:57 +0000)]
- Change the type of size_max to u_quad_t because its value is converted
  with vfs_scanopt(9) using the "%qu" format string.
- Limit the maximum value of size_max to (SIZE_MAX - PAGE_SIZE) to
  prevent overflow in howmany() macro.

PR: kern/141194
Approved by: trasz (mentor)
MFC after: 2 weeks