]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
13 years agoMFC
attilio [Sun, 29 May 2011 00:59:38 +0000 (00:59 +0000)]
MFC

13 years agoAdapt smusat(4) to use powermac_thermal. This provides automatic fan
nwhitehorn [Sat, 28 May 2011 22:11:22 +0000 (22:11 +0000)]
Adapt smusat(4) to use powermac_thermal. This provides automatic fan
management on dual- and quad-core Powermac G5s, and the last G5 iMacs.

13 years agoRequire an error instead of a timeout to decide the new-style fan
nwhitehorn [Sat, 28 May 2011 21:54:18 +0000 (21:54 +0000)]
Require an error instead of a timeout to decide the new-style fan
commands won't work. This prevents a busy system from making smu(4)
suddenly decide its fans use the old-style command set.

MFC after: 3 days

13 years agoFactor out the SMU fan management code into a new module (powermac_thermal)
nwhitehorn [Sat, 28 May 2011 21:13:22 +0000 (21:13 +0000)]
Factor out the SMU fan management code into a new module (powermac_thermal)
that will connect all of the various sensors and fan control modules on
Apple hardware with software-controlled fans (e.g. all G5 systems).

MFC after: 1 month

13 years agoo Determine the number of LAWs in a way the is future proof. Only the
marcel [Sat, 28 May 2011 19:14:16 +0000 (19:14 +0000)]
o   Determine the number of LAWs in a way the is future proof. Only the
    MPC8555(E) has 8 LAWs, so don't make that the default case. Current
    processors have 12 LAWs so use that as the default instead.
o   Determine the target ID of the PCI/PCI-X and PCI-E controllers in
    a way that's more future proof. There's almost a perfect mapping
    from HC register offset to target ID, so use that as the default.
    Handle the MPC8548(E) specially, since it has a non-standard target
    ID for the PCI-E controller. Don't worry about whether the processor
    implements the target ID here, because we should not get called for
    PCI/PCI-X or PCI-E host controllers that don't exist.

13 years agoRemove unused defines. They're distracting...
marcel [Sat, 28 May 2011 16:30:24 +0000 (16:30 +0000)]
Remove unused defines. They're distracting...

13 years agoFix shell-based partitioning.
nwhitehorn [Sat, 28 May 2011 16:15:46 +0000 (16:15 +0000)]
Fix shell-based partitioning.

13 years agoFix AR9287 operation when >1 TX chain is enabled.
adrian [Sat, 28 May 2011 15:43:56 +0000 (15:43 +0000)]
Fix AR9287 operation when >1 TX chain is enabled.

I didn't pick this up with the initial commit because I was only testing
with 11bg.

13 years agoUpdate the manual page to reflect the new 32K/4K defaults.
mckusick [Sat, 28 May 2011 15:14:50 +0000 (15:14 +0000)]
Update the manual page to reflect the new 32K/4K defaults.

Reminded by: Ivan Voras

13 years agoDue to a lag in updating the fs_pendinginodes count, we cannot depend
mckusick [Sat, 28 May 2011 15:07:29 +0000 (15:07 +0000)]
Due to a lag in updating the fs_pendinginodes count, we cannot depend
on it to decide whether we should try to reclaim inodes when we run
short.

Discovered by: Peter Holm

13 years agoprintf(1): Document that %c and precision for %b/%s use bytes, not chars.
jilles [Sat, 28 May 2011 14:32:47 +0000 (14:32 +0000)]
printf(1): Document that %c and precision for %b/%s use bytes, not chars.

This means these features do not work as expected with multibyte characters.

This perhaps less than ideal behaviour matches printf(3) and is specified by
POSIX.

13 years agoprintf: Allow multibyte characters for '<char> form, avoid negative codes.
jilles [Sat, 28 May 2011 11:37:47 +0000 (11:37 +0000)]
printf: Allow multibyte characters for '<char> form, avoid negative codes.

Examples:
  LC_ALL=en_US.UTF-8 printf '%d\n' $(printf \'\\303\\244)
  LC_ALL=en_US.ISO8859-1 printf '%d\n' $(printf \'\\344)
Both of these should print 228.

Like some other shells, incomplete or invalid multibyte characters yield the
value of the first byte without a warning.

Note that there is no general way to go back from the character code to the
character.

13 years agoNew boot loader menus from Devin Teske.
julian [Sat, 28 May 2011 08:50:38 +0000 (08:50 +0000)]
New boot loader menus from Devin Teske.
Discussed on hackers and recommended for inclusion into 9.0 at the devsummit.
All support email to devin   dteske at vicor dot ignoreme dot com .

Submitted by: dteske at vicor dot ignoreme dot com
Reviewed by: me and many others

13 years agoBetter support different kernel hand-offs. When loaded directly
marcel [Sat, 28 May 2011 04:10:44 +0000 (04:10 +0000)]
Better support different kernel hand-offs. When loaded directly
from U-Boot, the kernel is passed a standard argc/argv pair.
The Juniper loader passes the metadata pointer as the second
argument and passes 0 in the first. The FreeBSD loader passes
the metadata pointer in the first argument.

As such, have locore preserve the first 2 arguments in registers
r30 & r31. Change e500_init() to accept these arguments. Don't
pass global offsets (i.e. kernel_text and _end) as arguments to
e500_init(). We can reference those directly.

Rename e500_init() to booke_init() now that we're changing the
prototype.

In booke_init(), "decode" arg1 and arg2 to obtain the metadata
pointer correctly. For the U-Boot case, clear SBSS and BSS and
bank on having a static FDT for now. This allows loading the
ELF kernel and jumping to the entry point without trampoline.

13 years agoUpgrade to 9.6-ESV-R4-P1, which address the following issues:
dougb [Sat, 28 May 2011 00:21:28 +0000 (00:21 +0000)]
Upgrade to 9.6-ESV-R4-P1, which address the following issues:

1. Very large RRSIG RRsets included in a negative cache can trigger
an assertion failure that will crash named (BIND 9 DNS) due to an
off-by-one error in a buffer size check.

This bug affects all resolving name servers, whether DNSSEC validation
is enabled or not, on all BIND versions prior to today. There is a
possibility of malicious exploitation of this bug by remote users.

2. Named could fail to validate zones listed in a DLV that validated
insecure without using DLV and had DS records in the parent zone.

Add a patch provided by ru@ and confirmed by ISC to fix a crash at
shutdown time when a SIG(0) key is being used.

13 years agoo The P1020(E) & P2020(E) also have two cores. This conditional has
marcel [Fri, 27 May 2011 23:18:41 +0000 (23:18 +0000)]
o   The P1020(E) & P2020(E) also have two cores. This conditional has
    a tendency to grow unwieldy so we may want to revisit this in due
    time.
o   Simplify the CPU reset function by writing to the reset control
    register irrespective of whether the CPU has one and automatically
    falling back to the debug control register if we didn't reset the
    CPU. The side-effect is that we now properly reset future processors
    without first having to add the system version to the list.

13 years agoWire the kernel using TLB1 entry 0 rather than entry 1. A more recent
marcel [Fri, 27 May 2011 23:09:12 +0000 (23:09 +0000)]
Wire the kernel using TLB1 entry 0 rather than entry 1. A more recent
U-Boot as found on the P1020RDB doesn't like it when we use entry 1
(for some reason) whereas an older U-Boot doesn't mind if we use entry
0. If anything else, this simplifies the code a bit.

13 years agofind: If a part of an expression is unknown, do not call it an option.
jilles [Fri, 27 May 2011 22:14:49 +0000 (22:14 +0000)]
find: If a part of an expression is unknown, do not call it an option.

Although most of the primaries and operators start with "-", they are not
options.

Examples:
  find . -xyz
  find . -name xyz -or bad

MFC after: 1 week

13 years agoFix the new NFS client so that it handles NFSv4 state
rmacklem [Fri, 27 May 2011 22:05:10 +0000 (22:05 +0000)]
Fix the new NFS client so that it handles NFSv4 state
correctly during a forced dismount. This required that
the exclusive and shared (refcnt) sleep lock functions check
for MNTK_UMOUNTF before sleeping, so that they won't block
while nfscl_umount() is getting rid of the state. As
such, a "struct mount *" argument was added to the locking
functions. I believe the only remaining case where a forced
dismount can get hung in the kernel is when a thread is
already attempting to do a TCP connect to a dead server
when the krpc client structure called nr_client is NULL.
This will only happen just after a "mount -u" with options
that force a new TCP connection is done, so it shouldn't
be a problem in practice.

MFC after: 2 weeks

13 years agosh: Remove the "exp" builtin.
jilles [Fri, 27 May 2011 20:53:07 +0000 (20:53 +0000)]
sh: Remove the "exp" builtin.

The "exp" builtin is undocumented, non-standard and not very useful.

If exp's return value is not used, something like
VAR=$(exp EXPRESSION)
is equivalent to
VAR=$((EXPRESSION))
except that errors in the expression are fatal and quoting special
characters is not needed in the latter case.

If exp's return value is used, something like
if exp EXPRESSION >/dev/null
can be replaced by
if [ $((EXPRESSION)) -ne 0 ]
with similar differences.

The exp-run showed that "let" is close enough to bash's and ksh's builtin
that removing it would break a few ports. Therefore, "let" remains in 9.x.

PR: bin/104432
Exp-run done by: pav (with some other sh(1) changes)

13 years agosh: Correct criterion for using CDPATH in cd.
jilles [Fri, 27 May 2011 20:01:46 +0000 (20:01 +0000)]
sh: Correct criterion for using CDPATH in cd.

CDPATH should be ignored not only for pathnames starting with '/' but also
for pathnames whose first component is '.' or '..'.

The man page already describes this behaviour.

13 years agoRemove definitions for RACCT_FSIZE and RACCT_SBSIZE - these two are rather
trasz [Fri, 27 May 2011 19:57:58 +0000 (19:57 +0000)]
Remove definitions for RACCT_FSIZE and RACCT_SBSIZE - these two are rather
performance-sensitive and not that useful, so I won't be merging them
before 9.0.

13 years agosh: Add simple CDPATH test.
jilles [Fri, 27 May 2011 19:36:07 +0000 (19:36 +0000)]
sh: Add simple CDPATH test.

13 years agoWhitespace fixes
kevlo [Fri, 27 May 2011 16:17:35 +0000 (16:17 +0000)]
Whitespace fixes

Reviewed by: jpaetzel

13 years agoMFC
attilio [Fri, 27 May 2011 16:09:10 +0000 (16:09 +0000)]
MFC

13 years agoStyle fix: cast to size_t rather than u_long when comparing to sizeof()
attilio [Fri, 27 May 2011 16:01:51 +0000 (16:01 +0000)]
Style fix: cast to size_t rather than u_long when comparing to sizeof()
rets.

Requested by: kib

13 years agosh: Various updates to the TOUR document.
jilles [Fri, 27 May 2011 16:00:37 +0000 (16:00 +0000)]
sh: Various updates to the TOUR document.

13 years agosh: Fix unquoted $@/$* if IFS=''.
jilles [Fri, 27 May 2011 15:56:13 +0000 (15:56 +0000)]
sh: Fix unquoted $@/$* if IFS=''.

If IFS is null, unquoted $@/$* should still expand to separate words.
This differs from quoted $@ (which does not depend on IFS) in that pathname
generation is performed and empty words are removed.

13 years agoIn the near future cpuset_t objects in struct pcpu will be axed out, but
attilio [Fri, 27 May 2011 15:50:14 +0000 (15:50 +0000)]
In the near future cpuset_t objects in struct pcpu will be axed out, but
as long as this does not happen, we need to fix interfaces to userland
in order to not break run-time accesses to the structure.

Reviwed by: kib
Tested by: pluknet

13 years agoAdd example how to create MBR and BSD schemes and install boot code.
ae [Fri, 27 May 2011 15:29:39 +0000 (15:29 +0000)]
Add example how to create MBR and BSD schemes and install boot code.

13 years agoSynchronize manpage's synopsis with program's usage. Since -l and -r
ae [Fri, 27 May 2011 14:30:13 +0000 (14:30 +0000)]
Synchronize manpage's synopsis with program's usage. Since -l and -r
keys are mutually exclusive for the `gpart show` command, then mark
them so.

Requested by: ru

13 years ago- Move some internal functions to the private namespace
gabor [Fri, 27 May 2011 10:40:34 +0000 (10:40 +0000)]
- Move some internal functions to the private namespace

Submitted by: kan
Approved by: delphij (mentor)

13 years agoSilence warnings about unsupoorted value types.
pjd [Fri, 27 May 2011 08:34:31 +0000 (08:34 +0000)]
Silence warnings about unsupoorted value types.

MFC after: 2 weeks

13 years agoSome partitioning tools may have a different opinion about disk
ae [Fri, 27 May 2011 06:37:42 +0000 (06:37 +0000)]
Some partitioning tools may have a different opinion about disk
geometry and partitions may start from withing the first track.
If we found such partitions, then do not reserve space of the
first track, only first sector.

13 years agoo Swap the SVR numbers for MPC8533 & MPC8533E
marcel [Fri, 27 May 2011 05:58:10 +0000 (05:58 +0000)]
o   Swap the SVR numbers for MPC8533 & MPC8533E
o   Add SVR defines for P1011(E), P1020(E), P2010(E) & P2020(E)

13 years agoChange new constant names to ones used by OpenSolaris.
mav [Fri, 27 May 2011 03:44:47 +0000 (03:44 +0000)]
Change new constant names to ones used by OpenSolaris.

13 years agoOops, fix typo in r222336.
mav [Fri, 27 May 2011 03:30:23 +0000 (03:30 +0000)]
Oops, fix typo in r222336.

13 years agoAdd ses/srcs/eltsub.h to the list of files to be installed.
mav [Fri, 27 May 2011 03:27:28 +0000 (03:27 +0000)]
Add ses/srcs/eltsub.h to the list of files to be installed.
It is required for user-level SES example tools build.

13 years agoAdd names for few more SES element types according SES-2 specification.
mav [Fri, 27 May 2011 03:23:39 +0000 (03:23 +0000)]
Add names for few more SES element types according SES-2 specification.

13 years agoThe check for whether a block is going to be claimed by a snapshot
mckusick [Thu, 26 May 2011 23:56:58 +0000 (23:56 +0000)]
The check for whether a block is going to be claimed by a snapshot
needs to happen before we notify the underlying layer that it is
being freed.

13 years agoIn ipoib_cm_handle_rx_wc(): Count incoming packets and
delphij [Thu, 26 May 2011 22:29:43 +0000 (22:29 +0000)]
In ipoib_cm_handle_rx_wc(): Count incoming packets and
bytes toward incoming counters.

Reviewed by: jeff

13 years agoAdd a check for MNTK_UNMOUNTF at the beginning of nfs_sync()
rmacklem [Thu, 26 May 2011 22:05:35 +0000 (22:05 +0000)]
Add a check for MNTK_UNMOUNTF at the beginning of nfs_sync()
in the new NFS client so that a forced dismount doesn't
get stuck in the VFS_SYNC() call that happens before
VFS_UNMOUNT() in dounmount().
Additional changes are needed before forced dismounts will work.

MFC after: 2 weeks

13 years agoFor Timedia multiport serial adapters, the first two ports use a SUN1889
jhb [Thu, 26 May 2011 20:54:45 +0000 (20:54 +0000)]
For Timedia multiport serial adapters, the first two ports use a SUN1889
which uses a non-standard clock (* 8) while any additional ports use
SUN1699 chips which use a standard clock.

Tested by: N.J. Mann   njm of njm me uk
MFC after: 1 week

13 years agoDon't assume we have a valid bootinfo pointer.
marcel [Thu, 26 May 2011 20:47:05 +0000 (20:47 +0000)]
Don't assume we have a valid bootinfo pointer.

13 years agoFix tftp_log() usage.
rodrigc [Thu, 26 May 2011 20:32:33 +0000 (20:32 +0000)]
Fix tftp_log() usage.

13 years agoIntroduce AR9287 support to the FreeBSD HAL.
adrian [Thu, 26 May 2011 20:31:08 +0000 (20:31 +0000)]
Introduce AR9287 support to the FreeBSD HAL.

13 years agoFix a macro name - it's currently unused in this file however, but
adrian [Thu, 26 May 2011 20:22:10 +0000 (20:22 +0000)]
Fix a macro name - it's currently unused in this file however, but
keep it consistent with ar9280.c.

13 years agoRevert this erroneous commit and re-disable the AR9285 combined antenna
adrian [Thu, 26 May 2011 20:17:59 +0000 (20:17 +0000)]
Revert this erroneous commit and re-disable the AR9285 combined antenna
diversity.

13 years agoInclude an EEPROM dump program for the AR9287 EEPROM format.
adrian [Thu, 26 May 2011 19:49:32 +0000 (19:49 +0000)]
Include an EEPROM dump program for the AR9287 EEPROM format.

13 years agoFix support for RACCT_CORE by merging forgotten file.
trasz [Thu, 26 May 2011 18:54:07 +0000 (18:54 +0000)]
Fix support for RACCT_CORE by merging forgotten file.

13 years agoRaise the default blocksize for UFS/FFS filesystems from
mckusick [Thu, 26 May 2011 18:22:49 +0000 (18:22 +0000)]
Raise the default blocksize for UFS/FFS filesystems from
16K to 32K and the default fragment size from 2K to 4K.

The rational is that most disks are now running with 4K
sectors.  While they can (slowly) simulate 512-byte sectors
by doing a read-modify-write, it is desirable to avoid this
functionality.  By raising the minimum filesystem allocation
to 4K, the filesystem will never trigger the small sector
emulation.

Also, the growth of disk sizes has lead us to double the
default block size about every ten years.  The rise from 8K
to 16K blocks was done in 2001.  So, by the 10-year metric,
the time has come for 32K blocks.

Discussed at: May 2011 BSDCan Developer Summit
Reference: http://wiki.freebsd.org/201105DevSummit/FileSystems

13 years agoMFC
attilio [Thu, 26 May 2011 17:38:00 +0000 (17:38 +0000)]
MFC

13 years agoIgnore MCR[6] during the probe to fix a false negative. Bit 6 of the
marcel [Thu, 26 May 2011 17:02:56 +0000 (17:02 +0000)]
Ignore MCR[6] during the probe to fix a false negative. Bit 6 of the
MCR register on the Sunix Sun1699 chip tends to be set but doesn't
seem to have a function. That is, FreeBSD just works (provided the
correct RCLK is used) regardless.

PR: kern/129663
Diagnostics: Eygene Ryabinkin <rea-fbsd at codelabs.ru>
MFC after: 3 days

13 years agoRemove the three-chain scaled power check for the AR9287 - it isn't
adrian [Thu, 26 May 2011 16:59:42 +0000 (16:59 +0000)]
Remove the three-chain scaled power check for the AR9287 - it isn't
needed.

13 years agoMake sure only two chains are calibrated for the AR9287.
adrian [Thu, 26 May 2011 16:55:44 +0000 (16:55 +0000)]
Make sure only two chains are calibrated for the AR9287.

13 years agoAdd some open-loop TX power debugging for AR9287.
adrian [Thu, 26 May 2011 16:52:37 +0000 (16:52 +0000)]
Add some open-loop TX power debugging for AR9287.

13 years agoClose a race between libzfs and mountd when updating NFS exports.
will [Thu, 26 May 2011 16:27:00 +0000 (16:27 +0000)]
Close a race between libzfs and mountd when updating NFS exports.
- Flush the file descriptor for the new ZFS exports file before
  sending a SIGHUP to mountd.

Reviewed by: pjd
Approved by: ken
MFC after: 3 days

13 years agoBring over the AR5416 per-rate TX power code, modified to use the
adrian [Thu, 26 May 2011 15:55:27 +0000 (15:55 +0000)]
Bring over the AR5416 per-rate TX power code, modified to use the
AR9287 EEPROM layout.

The AR9287 only supports 2ghz, so I've removed the 5ghz code (but left
the 5ghz edge flags in there for now) and hard-coded the 2ghz-only
path.

Whilst I'm there, fix a typo (ar9285->ar9287.)

This meets basic TX throughput testing - iperf TX tests == 27-28mbit in 11g,
matching the rest of my 11g kit.

13 years agoFlesh out ar9287SetTransmitPower() based on the AR9285 routine.
adrian [Thu, 26 May 2011 15:01:37 +0000 (15:01 +0000)]
Flesh out ar9287SetTransmitPower() based on the AR9285 routine.

Hard-code the per-rate TX power at 5dBm for now so testing can be done.

This passes initial TX testing in 11g mode (but, obviously, at 5dBm.)

13 years agoAdd a missing isync.
nwhitehorn [Thu, 26 May 2011 14:34:22 +0000 (14:34 +0000)]
Add a missing isync.

13 years agoFlesh out the TX power calibration for the AR9287.
adrian [Thu, 26 May 2011 14:29:05 +0000 (14:29 +0000)]
Flesh out the TX power calibration for the AR9287.

I'm assuming for now that the AR9287 is only open-loop TX power control
(as mine is) so I've hard-coded the attach path to fail if the NIC is
not open-loop.

This greatly simplifies the TX calibration path and the amount of code
which needs to be ported over.

This still isn't complete - the rate calculation code still needs to be
ported and it all needs to be glued together.

Obtained from: Linux ath9k

13 years agoAdded myself as src committer.
ray [Thu, 26 May 2011 13:54:07 +0000 (13:54 +0000)]
Added myself as src committer.

Approved by: adrian (mentor)

13 years agoAdd Marvell 88SE9172 chip PCI ID.
mav [Thu, 26 May 2011 10:10:10 +0000 (10:10 +0000)]
Add Marvell 88SE9172 chip PCI ID.

13 years agoAdd the AR9287 chip identification string.
adrian [Thu, 26 May 2011 09:27:58 +0000 (09:27 +0000)]
Add the AR9287 chip identification string.

13 years agoMarvell 88SE91xx controllers are known to report soft-reset completion
mav [Thu, 26 May 2011 09:23:01 +0000 (09:23 +0000)]
Marvell 88SE91xx controllers are known to report soft-reset completion
without waiting for device readiness (or at least not updating FIS receive
area in time). To workaround that, special quirk was added earlier to wait
for the FIS receive area update. But it was found that under same PCI ID
0x91231b4b and revision 0x11 there are two completely different chip
versions (firmware?): HBA and RAID. The problem is that RAID version in
some cases, such as hot-plug, does not update FIS receive area at all!

To workaround that, differentiate the chip versions by their capabilities,
and, if RAID version found, skip FIS receive area update waiting and read
device signature from the PxSIG register instead. This method doesn't work
for HBA version when PMP attached, so keep using previous workaround there.

13 years agoFix a bad merge from a previous commit.
adrian [Thu, 26 May 2011 09:22:59 +0000 (09:22 +0000)]
Fix a bad merge from a previous commit.

13 years agoMerlin -> Kiwi
adrian [Thu, 26 May 2011 09:16:09 +0000 (09:16 +0000)]
Merlin -> Kiwi

13 years agoBring over my AR9287 work in progress.
adrian [Thu, 26 May 2011 09:15:33 +0000 (09:15 +0000)]
Bring over my AR9287 work in progress.

It isn't linked into the build because it's missing the TX power
and PDADC programming code.

This code is mostly based on the ath9k codebase, compared against
the Atheros codebase as appropriate.

What's implemented:

* probe/attach
* EEPROM board value programming
* RX initial calibration
* radio channel programming
* general MAC / baseband setup
* async fifo setup
* open-loop tx power calibration

What's missing before it can be enabled by default:

* TX power / calibration setting code
* closed-loop tx power calibration routines
* TSF2 handling
* generic timer support from ath9k

Obtained from: Atheros, ath9k

13 years agoAR9287 prep work:
adrian [Thu, 26 May 2011 08:35:47 +0000 (08:35 +0000)]
AR9287 prep work:

* Add PCI/PCIE devids
* Add AR9287/Kiwi version check macros
* AR_SREV_9287 -> AR_SREV_KIWI

Obtained from: Atheros, ath9k

13 years agoAdd temp sense to the EEPROM variable list;
adrian [Thu, 26 May 2011 08:20:14 +0000 (08:20 +0000)]
Add temp sense to the EEPROM variable list;
Export the temperature sense variables to ah_eeprom_9287.c

13 years agoAdd better names for the Intel HDMI audio codecs.
mav [Thu, 26 May 2011 06:43:10 +0000 (06:43 +0000)]
Add better names for the Intel HDMI audio codecs.

13 years ago+ Tighten up (and simplify) the pass_cmd_vars_1 "variable definition arrived
obrien [Wed, 25 May 2011 23:33:49 +0000 (23:33 +0000)]
+ Tighten up (and simplify) the pass_cmd_vars_1 "variable definition arrived
  from the calling make" test.
+ Be more tolerant of newlines in the plus_flag "supports the '+' flag" test.

13 years agosh: Show errno messages in cd.
jilles [Wed, 25 May 2011 21:38:16 +0000 (21:38 +0000)]
sh: Show errno messages in cd.

13 years agoAdd some missing mutex locking to the new NFS client.
rmacklem [Wed, 25 May 2011 21:17:53 +0000 (21:17 +0000)]
Add some missing mutex locking to the new NFS client.

MFC after: 2 weeks

13 years agoBump the date of the man page to the date of the actual commit.
bcr [Wed, 25 May 2011 21:04:11 +0000 (21:04 +0000)]
Bump the date of the man page to the date of the actual commit.

Noticed by: brix

13 years agoFix the new NFS client so that it correctly sets the "must_commit"
rmacklem [Wed, 25 May 2011 20:53:08 +0000 (20:53 +0000)]
Fix the new NFS client so that it correctly sets the "must_commit"
argument for a write RPC when it succeeds for the first one and
fails for a subsequent RPC within the same call to the function.
This makes it compatible with the old NFS client for this case.

MFC after: 2 weeks

13 years agoDocument the device name change from gpioctl to gpioc in the
bcr [Wed, 25 May 2011 20:25:13 +0000 (20:25 +0000)]
Document the device name change from gpioctl to gpioc in the
man page.

PR: docs/157075
Submitted by: brix
Reviewed by: gonzo

13 years agoFix a regression introduced with previous changeset: if output is stdout,
delphij [Wed, 25 May 2011 18:04:11 +0000 (18:04 +0000)]
Fix a regression introduced with previous changeset: if output is stdout,
do not check for symbolic link.

13 years ago[mdoc] Fixed .Dt call.
ru [Wed, 25 May 2011 14:13:53 +0000 (14:13 +0000)]
[mdoc] Fixed .Dt call.

13 years agoAccording to SATA specification, when Serial ATA Enclosure Management Bridge
mav [Wed, 25 May 2011 13:55:49 +0000 (13:55 +0000)]
According to SATA specification, when Serial ATA Enclosure Management Bridge
(SEMB) is unable to communicate to Storage Enclosure Processor (SEP), in
response to hard and soft resets it should among other things return value
0x7F in Status register. The weird side is that it means DRQ bit set, which
tells that reset request is not completed. It would be fine if SEMB was the
only device on port. But if SEMB connected to PMP or built into it, it may
block access to other devices sharing same SATA port.

Make some tunings/fixes to soft-reset handling to workaround the issue:
 - ahci(4): request CLO on the port after soft reset to ignore DRQ bit;
 - siis(4): gracefully reinitialize port after soft reset timeout (hardware
doesn't detect reset request completion in this case);
 - mvs(4): if PMP is used, send dummy soft-reset to the PMP port to make it
clear DRQ bit for us.

For now this makes quirks in ata_pmp.c, hiding SEMB ports of SiI3726/SiI4726
PMPs, less important. Further, if hardware permit, I hope to implement real
SEMB support.

13 years agoPrevent non-aligned reading from provider while tasting. Reject
ae [Wed, 25 May 2011 11:14:26 +0000 (11:14 +0000)]
Prevent non-aligned reading from provider while tasting. Reject
providers with unsupported sectorsize.

Reported by: Joerg Wunsch
MFC after: 1 week

13 years agoBring back r222275. runfw(4) will statically link in rt2870.fw.uu
kevlo [Wed, 25 May 2011 10:04:13 +0000 (10:04 +0000)]
Bring back r222275. runfw(4) will statically link in rt2870.fw.uu
to the kernel, though I have MODULES_OVERRIDE="" in GENERIC.

Spotted by: thompsa

13 years agoDo not truncate available disk space to the closest track boundary.
ae [Wed, 25 May 2011 09:45:13 +0000 (09:45 +0000)]
Do not truncate available disk space to the closest track boundary.

13 years agoDo not truncate available disk space to the closest track boundary.
ae [Wed, 25 May 2011 09:38:12 +0000 (09:38 +0000)]
Do not truncate available disk space to the closest track boundary.

13 years agoDo not truncate available disk space to the closest track boundary.
ae [Wed, 25 May 2011 09:32:19 +0000 (09:32 +0000)]
Do not truncate available disk space to the closest track boundary.

13 years agoAdd a description to the checksum target about not only
bcr [Wed, 25 May 2011 08:42:01 +0000 (08:42 +0000)]
Add a description to the checksum target about not only
being able to verify, but also having the ability to
fetch distfiles that are missing or failed the checksum
calculation

PR: docs/138887
Submitted by: Radim Kolar (hsn at sendmail dot cz)
MFC after: 5 days

13 years agoThe current ANI capability information uses a different set of
adrian [Wed, 25 May 2011 07:34:49 +0000 (07:34 +0000)]
The current ANI capability information uses a different set of
values for the commands, compared to the internal command values
(HAL_ANI_CMD.)

My eventual aim is to make the HAL_ANI_CMD internal enum match
the public API and then remove all this messiness.

This now allows HAL_CAP_INTMIT users to use a public HAL_CAP_INTMIT_
enum rather than magic constants.

The only magic constants currently used by if_ath are "enable" and
"present". Some local tools of mine allow for direct, manual fiddling
of the ANI variables and I'll convert these to use the public enum API
before I commit them.

13 years agoTidy up the ANI API in preparation for looking to expose some more
adrian [Wed, 25 May 2011 07:19:19 +0000 (07:19 +0000)]
Tidy up the ANI API in preparation for looking to expose some more
of the ANI statistics and committing some tools which use these.

* Change HAL_ANI_* commands _back_ to be numerical, rather than a
  bitmap;
* modify access to the ANI control bitmap to convert a command to
  a bitmap;
* Fix the ANI noise immunity fiddling for CCK errors - it wasn't
  checking whether noise immunity was disabled or not.

13 years agorun(4) needs firmware loaded to work
kevlo [Wed, 25 May 2011 04:46:48 +0000 (04:46 +0000)]
run(4) needs firmware loaded to work

13 years agoBuild and install a BSD licensed grep.
obrien [Wed, 25 May 2011 01:04:12 +0000 (01:04 +0000)]
Build and install a BSD licensed grep.
If WITH_BSD_GREP is not set, it will be 'bsdgrep' and GNUgrep will be
'[ef]grep'.  Otherwise, BSD-grep will be the grep family, and GNUgrep
will be 'gnugrep'.

Discussed with: brooks

13 years agoAdd FEATURE() definitions for IPv4 and IPv6 so that we can use
bz [Wed, 25 May 2011 00:34:25 +0000 (00:34 +0000)]
Add FEATURE() definitions for IPv4 and IPv6 so that we can use
feature_present(3) to dynamically decide whether to use one or the
other family.

Reviewed by: gnn
Sponsored by: The FreeBSD Foundation
Sponsored by: iXsystems
MFC after: 10 days

13 years agostyle(9)
yongari [Tue, 24 May 2011 20:39:07 +0000 (20:39 +0000)]
style(9)

13 years agoDon't pass pointer to name buffer which is on the stack to another thread,
pjd [Tue, 24 May 2011 20:10:12 +0000 (20:10 +0000)]
Don't pass pointer to name buffer which is on the stack to another thread,
because the stack might be paged out once the other thread tries to use the
data. Instead, just allocate memory.

MFC after: 2 weeks

13 years agoDon't access task structure once we call task function.
pjd [Tue, 24 May 2011 20:07:15 +0000 (20:07 +0000)]
Don't access task structure once we call task function.
The task structure might be no longer available.
This also allows to eliminates the need for two tasks in the zio structure.

Submitted by: anonymous
MFC after: 2 weeks

13 years agoSilly spelling typos.
jhb [Tue, 24 May 2011 19:55:57 +0000 (19:55 +0000)]
Silly spelling typos.

Submitted by: "b. f."

13 years agoThe ANI control for the AR5416 and later chips was calling ar5212AniControl(),
adrian [Tue, 24 May 2011 18:25:40 +0000 (18:25 +0000)]
The ANI control for the AR5416 and later chips was calling ar5212AniControl(),
which did AR5212 specific initialisation. This would cause some slight
silliness when enabling/disabling ANI.

Just to be completely correct - and to ensure the phy error mask/RX filter
register isn't incorrectly played with - make the ANI control function a
method, have it set appropriately for AR5212/AR5416, and call that from the
ANI control interface.

13 years agoSimplify ALIGNDOWN macro.
ae [Tue, 24 May 2011 17:03:46 +0000 (17:03 +0000)]
Simplify ALIGNDOWN macro.

13 years agoFix calculation of alignment for odd values. Also do not change value
ae [Tue, 24 May 2011 16:49:34 +0000 (16:49 +0000)]
Fix calculation of alignment for odd values. Also do not change value
when it is already aligned.

13 years ago- Fix a misusage of cpuset_t objects
attilio [Tue, 24 May 2011 15:47:40 +0000 (15:47 +0000)]
- Fix a misusage of cpuset_t objects
- Fix a typo

Reported by: pluknet

13 years agoAssume the link to be dead if bit error rate (BER) parameter is set to 1.
zec [Tue, 24 May 2011 14:36:32 +0000 (14:36 +0000)]
Assume the link to be dead if bit error rate (BER) parameter is set to 1.
When a transition from link alive to link dead configuration or vice
versa occurs, notify any upstream and / or downstream peers using
NGM_FLOW messagges.

Link state notification using NGM_FLOW messages is modelled around
around already existing code in ng_ether.c.

MFC after: 3 days