]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years ago- Do not read and write RX configuration register multiple times.
jkim [Tue, 20 Jan 2009 19:58:01 +0000 (19:58 +0000)]
- Do not read and write RX configuration register multiple times.
- Always program RX configuration register from scratch instead of
doing read/modify/write.
- Rename re_setmulti() to re_set_rxmode() to be reflect reality.
- Simplify hash filter logic a little while I am here.

Reviewed by: yongari (early version)

15 years agoFix a number of (innocuous) warnings, and remove a useless test.
luigi [Tue, 20 Jan 2009 18:16:31 +0000 (18:16 +0000)]
Fix a number of (innocuous) warnings, and remove a useless test.
There are still several signed/unsigned warnings left, which
require a bit more study for a proper fix.

This file has grown beyond reasonable limits.

We really need to split it into separate components (ipv4, ipv6,
dummynet, nat, table, userland-kernel communication ...) so we can
make mainteinance easier.

MFC after: 1 weeks

15 years agoImplement MMCBR_IVAR_CAPS. It should better be implemented, or results
mav [Tue, 20 Jan 2009 17:36:58 +0000 (17:36 +0000)]
Implement MMCBR_IVAR_CAPS. It should better be implemented, or results
can be unpredictable.

PR: arm/128987

15 years agoRemove inlining of functions that are used mostly in different object files.
rdivacky [Tue, 20 Jan 2009 17:15:12 +0000 (17:15 +0000)]
Remove inlining of functions that are used mostly in different object files.
This gets rid of gnu89 style inlining. Also silence gcc by assigning two
variables NULL. This lets use to remove NO_WERROR.

Approved by: kib (mentor)
Approved by: harti

15 years agoAdd a comment explaining why the "bufwait" / "dirhash" LOR reported by
jhb [Tue, 20 Jan 2009 16:35:34 +0000 (16:35 +0000)]
Add a comment explaining why the "bufwait" / "dirhash" LOR reported by
WITNESS will not actually result in a deadlock.

Discussed with: kib
MFC after: 1 week

15 years agoFix a race condition in kiic(4) made possible by the way the device's STOP
nwhitehorn [Tue, 20 Jan 2009 14:09:12 +0000 (14:09 +0000)]
Fix a race condition in kiic(4) made possible by the way the device's STOP
condition is sent. We used to put the bus in the STOP state, but returned
without waiting for that to actually occur.

Submitted by: Marco Trillo

15 years agoSet the wrong softc size when defining the ofw_iicbus class. Change it to
nwhitehorn [Tue, 20 Jan 2009 14:06:30 +0000 (14:06 +0000)]
Set the wrong softc size when defining the ofw_iicbus class. Change it to
the correct value.

15 years ago - Permit timestamps to be as far as 2048 ticks apart before we complain
jeff [Tue, 20 Jan 2009 12:33:04 +0000 (12:33 +0000)]
 - Permit timestamps to be as far as 2048 ticks apart before we complain
   about invalid timestamps.  Nehalem CPUs seem to be synchronized but only
   within a fraction of a microsecond.
 - Make the Counter code more flexible to poor timestamps.  In general we
   now complain a lot but render as much as we can.
 - Change the scaler behavior so it works better with very long and very
   short traces.  We now set the maximum scale such that it properly
   displays the entire file by default and doesn't permit zooming out
   beyond the file.  This improves other awkward navigation behavior.
   The interval is now set very small which can't be achieved by simply
   dragging the mouse.  Clicking to the left of or right of the scaler bar
   will produce increments of a single, very small, interval now.

Sponsored by:   Nokia

15 years agoThe context switch to the 32bit binary does not properly restore
kib [Tue, 20 Jan 2009 12:07:49 +0000 (12:07 +0000)]
The context switch to the 32bit binary does not properly restore
the fsbase value. The switch loads the fs segment register, that
invalidates the value in fsbase msr, thus value in %r9 can not be
considered the current value for fsbase anymore.

Unconditionally reload fsbase when switching to 32bit binary.

PR: 130526
MFC after: 3 weeks

15 years agoProperly implement the VT100 SCS sequences in xterm-mode.
ed [Tue, 20 Jan 2009 11:34:28 +0000 (11:34 +0000)]
Properly implement the VT100 SCS sequences in xterm-mode.

Even though VT100-like devices can display non-ASCII characters, they do
not use an 8-bit character set. Special escape sequences allow the VT100
to switch character maps. The special graphics character set stores the
box drawing characters, starting at 0x60, ending at 0x7e. This means
we now pass the character map tests in vttest, even the save/restore
cursor test, combined with character maps. dialog(1) also works a lot
better now.

This commit also includes some other minor fixes:

- Default to 24 lines in teken_demo when using xterm emulation.
- Make white foreground and background work in teken_demo.

15 years agoWhen extending inode size, we call vnode_pager_setsize(), to have a
kib [Tue, 20 Jan 2009 11:30:22 +0000 (11:30 +0000)]
When extending inode size, we call vnode_pager_setsize(), to have a
address space where to put vnode pages, and then call UFS_BALLOC(),
to actually allocate new block and map it. When UFS_BALLOC() returns
error, sometimes we forget to revert the vm object size increase,
allowing for the pages that are not backed by the logical disk blocks.

Revert vnode_pager_setsize() back when UFS_BALLOC() failed, for
ffs_truncate() and ffs_write().

PR: 129956
Reviewed by: ups
MFC after: 3 weeks

15 years agoFFS puts the extended attributes blocks at the negative blocks for the
kib [Tue, 20 Jan 2009 11:27:45 +0000 (11:27 +0000)]
FFS puts the extended attributes blocks at the negative blocks for the
vnode, from -1 down. When vinvalbuf(vp, V_ALT) is done for the vnode, it
incorrectly does vm_object_page_remove(0, 0), removing all pages from
the underlying vm object, not only the pages that back the extended
attributes data.

Change vinvalbuf() to not remove any pages from the object when
V_NORMAL or V_ALT are specified. Instead, the only in-tree caller
in ffs_inode.c:ffs_truncate() that specifies V_ALT explicitely
removes the corresponding page range. The V_NORMAL caller
does vnode_pager_setsize(vp, 0) immediately after the call to
vinvalbuf(V_NORMAL) already.

Reported by: csjp
Reviewed by: ups
MFC after: 3 weeks

15 years agoAdd a limit on namecache entries.
mckay [Tue, 20 Jan 2009 04:21:21 +0000 (04:21 +0000)]
Add a limit on namecache entries.

In normal operation, the number of cache entries is roughly equal to the
number of active vnodes.  However, when most of the recently accessed
vnodes have many hard links, the number of cache entries can be 32000
times as large, exhausting kernel memory and provoking a panic in
kmem_malloc().

MFC after: 2 weeks

15 years agoTone down warning about the quality of the NTFS VFS module. It appears that
sobomax [Tue, 20 Jan 2009 02:08:21 +0000 (02:08 +0000)]
Tone down warning about the quality of the NTFS VFS module. It appears that
not all developers share luigi opinion about quality of sysutils/fusefs-ntfs
compared to our kernel NTFS module.

15 years agoChange the probe priority for PCI and I2C generic bus modules from
nwhitehorn [Tue, 20 Jan 2009 00:05:43 +0000 (00:05 +0000)]
Change the probe priority for PCI and I2C generic bus modules from
numerical constants to BUS_PROBE_GENERIC.

Suggested by: jhb

15 years agoProvide a device description for macio-attached ATA cells.
nwhitehorn [Mon, 19 Jan 2009 23:25:18 +0000 (23:25 +0000)]
Provide a device description for macio-attached ATA cells.

15 years agoProperly return error code to the caller. This should fix the following
emax [Mon, 19 Jan 2009 22:06:35 +0000 (22:06 +0000)]
Properly return error code to the caller. This should fix the following
panic in ng_l2cap(4).

panic: ng_l2cap_l2ca_con_req: ubt0l2cap - could not find connection!

While i'm here get rid of few goto's.

MFC after: 1 week

15 years agoSet the pipe pointer before calling usbd_transfer() as its possible for the
thompsa [Mon, 19 Jan 2009 21:22:10 +0000 (21:22 +0000)]
Set the pipe pointer before calling usbd_transfer() as its possible for the
xfer callback to be invoked on error.

MFC after: 2 weeks

15 years agoAdd two more nVidia HDMI codec IDs.
mav [Mon, 19 Jan 2009 19:16:50 +0000 (19:16 +0000)]
Add two more nVidia HDMI codec IDs.

15 years agore-enable wi: was accidentally disabled in r1.502
sam [Mon, 19 Jan 2009 17:20:09 +0000 (17:20 +0000)]
re-enable wi: was accidentally disabled in r1.502

15 years agoMention removal of NTFS from GENERIC/amd64.
sobomax [Mon, 19 Jan 2009 17:00:42 +0000 (17:00 +0000)]
Mention removal of NTFS from GENERIC/amd64.

15 years agoMention the fact that the NTFS kernel support isn't
sobomax [Mon, 19 Jan 2009 16:19:53 +0000 (16:19 +0000)]
Mention the fact that the NTFS kernel support isn't
very well maintained and point user to sysutils/fusefs-ntfs, which
at the time of this writing seems to be a better alternative.

Suggested by: luigi
MFC after: 2 weeks

15 years agoIn the CAVEATS section mention the fact that the NTFS kernel support isn't
sobomax [Mon, 19 Jan 2009 15:59:05 +0000 (15:59 +0000)]
In the CAVEATS section mention the fact that the NTFS kernel support isn't
very well maintained and point user to sysutils/fusefs-ntfs, which
at the time of this writing seems to be a better alternative.

Suggested by: luigi

MFC after: 2 weeks

15 years agoTake NTFS option out to match i386 GENERIC.
sobomax [Mon, 19 Jan 2009 15:33:06 +0000 (15:33 +0000)]
Take NTFS option out to match i386 GENERIC.

Suggested by: phk, luigi

15 years agoasr(4) is not amd64-clean, not amr(4).
sobomax [Mon, 19 Jan 2009 08:51:20 +0000 (08:51 +0000)]
asr(4) is not amd64-clean, not amr(4).

Pointy hat to: myself
Submitted by: scottl

15 years agoComment amr(4) out - according to scottl it's not 64-bit clean.
sobomax [Mon, 19 Jan 2009 08:25:41 +0000 (08:25 +0000)]
Comment amr(4) out - according to scottl it's not 64-bit clean.

15 years agoWhitespace-only: reduce diff to the i386 GENERIC.
sobomax [Mon, 19 Jan 2009 07:18:32 +0000 (07:18 +0000)]
Whitespace-only: reduce diff to the i386 GENERIC.

15 years agoAdd asr(4) and stge(4) from i386 GENERIC. Both drivers compile on amd64 and
sobomax [Mon, 19 Jan 2009 07:10:11 +0000 (07:10 +0000)]
Add asr(4) and stge(4) from i386 GENERIC. Both drivers compile on amd64 and
there is no particular reason for them to be i386-only.

MFC after: 2 weeks

15 years ago- Add support for multibyte decimal_point encodings, e.g., U+066B.
das [Mon, 19 Jan 2009 06:19:51 +0000 (06:19 +0000)]
- Add support for multibyte decimal_point encodings, e.g., U+066B.
  A forthcoming gdtoa import is needed to make this fully work.
- Improve the way "nan(...)" is parsed.

15 years agoAdd support for multibyte decimal_point encodings, e.g., U+066B.
das [Mon, 19 Jan 2009 06:19:38 +0000 (06:19 +0000)]
Add support for multibyte decimal_point encodings, e.g., U+066B.

15 years agoSometimes RTL8168B seems to take long time to access GMII registers
yongari [Mon, 19 Jan 2009 02:31:27 +0000 (02:31 +0000)]
Sometimes RTL8168B seems to take long time to access GMII registers
in device attach phase. Double GMII register access timeout value
to fix the issue.

Reported by: wkoszek
Tested by: wkoszek

15 years agoAlso strip the machine arch from SRCDIR in case it is a cross build so
thompsa [Sun, 18 Jan 2009 23:21:04 +0000 (23:21 +0000)]
Also strip the machine arch from SRCDIR in case it is a cross build so
svnversion works.

15 years agoIf source mbuf chain consists of only one mbuf, use it directly as source
mav [Sun, 18 Jan 2009 21:09:34 +0000 (21:09 +0000)]
If source mbuf chain consists of only one mbuf, use it directly as source
buffer to avoid extra copying.

15 years agoTeach m_copyback() to use trailing space of the last mbuf in chain.
mav [Sun, 18 Jan 2009 20:19:55 +0000 (20:19 +0000)]
Teach m_copyback() to use trailing space of the last mbuf in chain.

15 years agoUse m_unshare()+m_copyback() instead of m_freem()+m_devget() to keep
mav [Sun, 18 Jan 2009 19:25:36 +0000 (19:25 +0000)]
Use m_unshare()+m_copyback() instead of m_freem()+m_devget() to keep
original mbuf chain headers. It can be less efficient in some cases, but it
looks better then mess of copying headers into the nonempty chain.

15 years ago- Eliminate warnings in debug print macros by explicitly converting all
stas [Sun, 18 Jan 2009 15:10:46 +0000 (15:10 +0000)]
- Eliminate warnings in debug print macros by explicitly converting all
  field to unsigned long.

15 years ago- Whitespace fixes.
stas [Sun, 18 Jan 2009 14:54:46 +0000 (14:54 +0000)]
- Whitespace fixes.
- s_bmask field doesn't exist.
- Use correct flags in debug printf.

15 years ago- Obtain inode sizes and location of the first inode based on the contents
stas [Sun, 18 Jan 2009 14:04:56 +0000 (14:04 +0000)]
- Obtain inode sizes and location of the first inode based on the contents
  of superblock rather than using hardcoded values. This fixes ext2fs on
  filesystems with inode sized other than 128.

Submitted by: Alex Lyashkov <Alexey.Lyashkov@Sun.COM> (based on)
MFC after: 2 weeks

15 years agoAdjust the padding of struct pcpu to r187357.
marius [Sun, 18 Jan 2009 13:04:38 +0000 (13:04 +0000)]
Adjust the padding of struct pcpu to r187357.

15 years agoRemove strict limitation on minimal multilink MRRU. RFC claims that MRRU
mav [Sun, 18 Jan 2009 12:03:43 +0000 (12:03 +0000)]
Remove strict limitation on minimal multilink MRRU. RFC claims that MRRU
of 1500 must be supported, but allows smaller values to be negotiated.
Enforce specified MRRU for outgoing frames.

MFC after: 2 weeks

15 years agoFix regression introduced in rev. 173124:
mav [Sun, 18 Jan 2009 11:38:35 +0000 (11:38 +0000)]
Fix regression introduced in rev. 173124:
0.0.0.0/1 is not the same as 0.0.0.0/0.

MFC after: 1 month

15 years agoDon't forget to mark the color translation array as const.
ed [Sun, 18 Jan 2009 09:44:33 +0000 (09:44 +0000)]
Don't forget to mark the color translation array as const.

Spotted by: Christoph Mallon <christoph mallon gmx de>

15 years agoCorrect an error in revision 1.170 of this file. When get_pv_entry() is
alc [Sun, 18 Jan 2009 08:00:55 +0000 (08:00 +0000)]
Correct an error in revision 1.170 of this file.  When get_pv_entry() is
forced to reclaim pv entries, the one pv entry that it returns should not
be freed.

15 years agoremove too noisy DIAGNOSTIC code
sam [Sun, 18 Jan 2009 07:20:02 +0000 (07:20 +0000)]
remove too noisy DIAGNOSTIC code

Reviewed by: qingli

15 years ago - Add summary information to the title once the file is parsed rather than
jeff [Sun, 18 Jan 2009 05:44:31 +0000 (05:44 +0000)]
 - Add summary information to the title once the file is parsed rather than
   printing it to the terminal.  Now only parse errors go to the terminal.
 - Speedup drawing by raising and lowering tags only once everything has
   been drawn.  Surprisingly, it now takes a little longer to parse than
   it does to draw.
 - Parameterize the layout with X_ and Y_ defines that determine the sizes
   of various things.
 - Remove unnecessary tags.

15 years agoJust put the ifnet pointer first in the softc like it needs to be rather than
thompsa [Sun, 18 Jan 2009 05:35:58 +0000 (05:35 +0000)]
Just put the ifnet pointer first in the softc like it needs to be rather than
adding another deliciously evil pointer.

15 years ago - Significantly speedup hiding and displaying multiple rows by writing an
jeff [Sun, 18 Jan 2009 04:49:01 +0000 (04:49 +0000)]
 - Significantly speedup hiding and displaying multiple rows by writing an
   optimized single pass function for each.  This reduces the number of
   tkinter calls required to the minimum.
 - Add a right-click context menu for sources.  Supported commands hide
   the source, hide the whole group the source is in, and bring up a stat
   window.
 - Add a source stat frame that gives an event frequency table as well as
   the total duration for each event type that has a duration.  This can
   be used to see, for example, the total time a thread spent running or
   blocked by a wchan or lock.

15 years agoAdd support for CMedia CMI120.
keramida [Sun, 18 Jan 2009 04:29:42 +0000 (04:29 +0000)]
Add support for CMedia CMI120.

PR: kern/123640
Approved by: ariff
MFC after: 2 weeks

15 years agoFix for my previous commit: color mapping is not 1:1.
ed [Sat, 17 Jan 2009 23:01:40 +0000 (23:01 +0000)]
Fix for my previous commit: color mapping is not 1:1.

Cons25 doesn't seem to use a straight 1:1 mapping to the ANSI colors,
but uses the same color numbers as at least used by syscons on i386. I
suspect if you change the definitions on a different architecture,
things may break? Not sure.

Add a small array to convert syscons-style color codes to ANSI
equivalents, which are used by libteken internally. I didn't notice this
bug, because I only tested my code with black, white and green, all of
them shared the same numbers.

15 years agoMake vidcontrol's color setting work again.
ed [Sat, 17 Jan 2009 22:53:53 +0000 (22:53 +0000)]
Make vidcontrol's color setting work again.

It turns out I forgot to implement two escape sequences that allows the
user to change the default foreground and background colors. I thought
they were implemented by syscons itself, but vidcontrol just generates
some escape sequences, which get interpreted by the terminal emulator.

Reported by: mgp (forums)

15 years agoFix typo
phk [Sat, 17 Jan 2009 21:31:47 +0000 (21:31 +0000)]
Fix typo

Spotted by: juli

15 years agoRelease the evil twin of nanobsd.sh: sysbuild.sh
phk [Sat, 17 Jan 2009 20:30:06 +0000 (20:30 +0000)]
Release the evil twin of nanobsd.sh: sysbuild.sh

quoth the README:

I have been running -current on my laptop since before FreeBSD 2.0 was
released and along the way developed this little trick to making the
task easier.

sysbuild.sh is a way to build a new FreeBSD system on a computer from
a specification, while leaving the current installation intact.

sysbuild.sh assume you have two partitions that can hold your rootfs
and can be booted, and roughly speaking, all it does is build a new
system into the one you don't use, from the one you do use.

A partition named /freebsd is assumed to be part of your layout, and
that is where the sources and ports will be found.

If you know how nanobsd works, you will find a lot of similarity.

15 years agoWhen f[w]printf() is called on an unbuffered file like stdout, it
das [Sat, 17 Jan 2009 18:57:12 +0000 (18:57 +0000)]
When f[w]printf() is called on an unbuffered file like stdout, it
sets up a fake buffered FILE and then effectively calls itself
recursively. Unfortunately, gcc doesn't know how to do tail call
elimination in this case, and actually makes things worse by
inlining __sbprintf(). This means that f[w]printf() to stderr was
allocating about 5k of stack on 64-bit platforms, much of which was
never used.

I've reorganized things to eliminate the waste. In addition to saving
some stack space, this improves performance in my tests by anywhere
from 5% to 17% (depending on the test) when -fstack-protector is
enabled. I found no statistically significant performance difference
when stack protection is turned off. (The tests redirected stderr to
/dev/null.)

15 years agoConvert battery capacity/rate from mA to mW only when summary is requested.
mav [Sat, 17 Jan 2009 17:40:44 +0000 (17:40 +0000)]
Convert battery capacity/rate from mA to mW only when summary is requested.
Unconditional conversion causes incorrect rate units reported by acpiconf.

MFC after: 2 weeks

15 years agoAllow experimental libteken features to be tested without changing code.
ed [Sat, 17 Jan 2009 16:37:13 +0000 (16:37 +0000)]
Allow experimental libteken features to be tested without changing code.

The teken library already supports UTF-8 handling and xterm emulation,
but we have reasons to disable this right now. Because we should make it
easy and interesting for people to experiment with these features, allow
them to be set in kernel configuration files.

Before this commit we had a flag called `TEKEN_CONS25' to enable
cons25-style emulation. I'm calling it the opposite now, `TEKEN_XTERM',
because we want to enable it in kernel configuration files explicitly.

Requested by: kib

15 years agoFix handling of pts(4) device names in comsat(8). Also catch fork() errors.
ed [Sat, 17 Jan 2009 15:56:38 +0000 (15:56 +0000)]
Fix handling of pts(4) device names in comsat(8). Also catch fork() errors.

Pseudo-terminals allocated with posix_openpt(2) will have more slashes
in their path names than comsat(8) allows, so allow slashes when the
character device name starts with "pts/".

This patch is loosely based on NetBSD's changes, revision 1.33. Because
it also included the changes to fork(), I imported them here as well.
Maybe we could import even more fixes from the other BSD's?

Original commit message from the NetBSD folks:

PR/30170: Markus W Kilbinger: src/libexec/comsat complains
about: '/' in "/dev/pts/1"

Reported by: Robert Huff <roberthuff rcn com>

15 years agos,unmount 8,umount 8, it is unmount(2) which I did not mean.
bz [Sat, 17 Jan 2009 14:52:26 +0000 (14:52 +0000)]
s,unmount 8,umount 8, it is unmount(2) which I did not mean.

Submitted by: pluknet@gmail.com
MFC after: 1 week

15 years agoThe FTP_TIMEOUT and HTTP_TIMEOUT environment variables were ignored because
des [Sat, 17 Jan 2009 13:34:56 +0000 (13:34 +0000)]
The FTP_TIMEOUT and HTTP_TIMEOUT environment variables were ignored because
T_secs already had a non-zero default.  Unbreak by moving the default to
ftp_timeout / http_timeout.

Submitted by: Christoph Mallon <christoph.mallon@gmx.de>
MFC after: 2 weeks

15 years agoBail on ENXIO, you won't get any further any way.
phk [Sat, 17 Jan 2009 11:57:32 +0000 (11:57 +0000)]
Bail on ENXIO, you won't get any further any way.

Submitted by: tobez

15 years ago - Add a new source configuration menu option that allows hiding and
jeff [Sat, 17 Jan 2009 11:19:15 +0000 (11:19 +0000)]
 - Add a new source configuration menu option that allows hiding and
   displaying sources.
 - Add functions to the main SchedGraph to facilitate source hiding.  The
   source is simply moved off screen and all other sources are moved to
   compensate.

15 years ago - Rewrite the parser to support the new generic schedgraph interface.
jeff [Sat, 17 Jan 2009 07:24:25 +0000 (07:24 +0000)]
 - Rewrite the parser to support the new generic schedgraph interface.
   This no longer requires any custom classes or parsers to support new
   event types.
 - Add an optional command line argument for specifying the clock frequency
   in ghz.  This is useful for traces that do not include KTR_SCHED.

Sponsored by: Nokia

 - Add support for sorting rows by clicking and dragging them to their new
   position.
 - Add support for configuring the cpu background colors.
 - Improve the scaling so a better center is maintained as you zoom.  This
   is not perfect due to precision loss with floats used in the window
   views.
 - Add new colors and a random assignment for unknown event types.  A table
   is used for known event types.  This is the only event specific
   information.

15 years ago - Implement generic macros for producing KTR records that are compatible
jeff [Sat, 17 Jan 2009 07:17:57 +0000 (07:17 +0000)]
 - Implement generic macros for producing KTR records that are compatible
   with src/tools/sched/schedgraph.py.  This allows developers to quickly
   create a graphical view of ktr data for any resource in the system.
 - Add sched_tdname() and the pcpu field 'name' for quickly and uniformly
   identifying records associated with a thread or cpu.
 - Reimplement the KTR_SCHED traces using the new generic facility.

Obtained from: attilio
Discussed with: jhb
Sponsored by: Nokia

15 years agoSimplify printf's inlined output buffering routines. On amd64, this
das [Sat, 17 Jan 2009 05:38:14 +0000 (05:38 +0000)]
Simplify printf's inlined output buffering routines. On amd64, this
reduces the code size by about 10% and improves performance slightly.

15 years agoDisable decoding of BARs by devices before we trash the value in the BAR
jhb [Fri, 16 Jan 2009 22:22:30 +0000 (22:22 +0000)]
Disable decoding of BARs by devices before we trash the value in the BAR
by writing all 1's to it to determine its length.  This fixes issues with
MCFG on at least some machines where a trashed BAR claimed subsequent
attempts at PCI config transactions because the addresses in the MCFG
window fell in the decoding range of the BAR.

In general it is a bad idea to leave the BARs enabled while we are
frobbing with them in this manner.

Sleuthing by:  tegge
MFC after:     1 week

15 years agoCorrect the sense of the "use32" terenary operator so that it uses
jhb [Fri, 16 Jan 2009 22:16:54 +0000 (22:16 +0000)]
Correct the sense of the "use32" terenary operator so that it uses
'lib32' for 32-bit binaries on amd64 and 'lib' for 64-bit binaries.
Remove an #ifdef __i386__ that defined use32 to the wrong value on
i386 to workaround the bug in the terenary operator.

MFC after: 1 month

15 years agoexport PSB frequencies
sam [Fri, 16 Jan 2009 20:44:45 +0000 (20:44 +0000)]
export PSB frequencies

15 years agorevert r187343
sam [Fri, 16 Jan 2009 20:42:43 +0000 (20:42 +0000)]
revert r187343

15 years agoadd FCC4 SKU to expose Public Safety Band (PSB) frequencies; this is
sam [Fri, 16 Jan 2009 20:40:50 +0000 (20:40 +0000)]
add FCC4 SKU to expose Public Safety Band (PSB) frequencies; this is
modeled after the Atheros SKU of the same name

15 years agoFix 1/2 and 1/4 width channel handling for non-GSM operation:
sam [Fri, 16 Jan 2009 20:38:42 +0000 (20:38 +0000)]
Fix 1/2 and 1/4 width channel handling for non-GSM operation:
o correct typo that caused random channel selection
o explicitly add 1/2 and 1/4 width channels because channel lookups match
  flags that include IEEE80211_CHANNEL_HALF and IEEE80211_CHANNEL_QUARTER

15 years agoUpdate URL.
marius [Fri, 16 Jan 2009 19:43:21 +0000 (19:43 +0000)]
Update URL.

15 years agoRemove a duplicate #include.
cognet [Fri, 16 Jan 2009 19:31:27 +0000 (19:31 +0000)]
Remove a duplicate #include.

15 years agoAdd some information in SYNOPSIS, HARDWARE and BUGS sections.
mav [Fri, 16 Jan 2009 19:28:25 +0000 (19:28 +0000)]
Add some information in SYNOPSIS, HARDWARE and BUGS sections.

15 years agoBe more Solaris-friendly.
pjd [Fri, 16 Jan 2009 18:09:49 +0000 (18:09 +0000)]
Be more Solaris-friendly.

Submitted by: Milan Cermak <Milan.Cermak@Sun.COM>

15 years agoAllow jot(1) regression tests to be checked out on Windows filesystems.
ed [Fri, 16 Jan 2009 15:47:35 +0000 (15:47 +0000)]
Allow jot(1) regression tests to be checked out on Windows filesystems.

The jot(1) regression tests directory contained two tests named `wx' and
`wX', which doesn't work on case insensitive filesystems. Rename `wX' to
`wX1'.

MFC after: 1 month

15 years agoRename mail(1)'s aux.c to util.c.
ed [Fri, 16 Jan 2009 15:00:30 +0000 (15:00 +0000)]
Rename mail(1)'s aux.c to util.c.

The name `aux' is reserved on Windows file systems. aux.c in the mail(1)
directory contains some random utility functions. I'm renaming this file
to util.c to make it possible to check out this directory on Windows.

MFC after: 1 month

15 years agoRemove unused files from the digi(4) driver.
ed [Fri, 16 Jan 2009 14:45:48 +0000 (14:45 +0000)]
Remove unused files from the digi(4) driver.

The digi(4) driver directory contains some files that cannot be checked
out on Windows filesystems. This isn't a big deal, but the files aren't
used anyway.

There are still some other places where checkouts on Windows don't work,
such as VFS_MOUNT.9/vfs_mount.9. This should already be a small
improvement.

MFC after: 1 month

15 years agoThe RTF_LLINFO was revived unconditionally, but within the kernel the
qingli [Fri, 16 Jan 2009 09:01:45 +0000 (09:01 +0000)]
The RTF_LLINFO was revived unconditionally, but within the kernel the
check on the sysctl argument value being RTF_LLINFO is conditioned on
the COMPAT_ROUTE_FLAGS kernel option. This mismatch caused the L2
table retrieval failure, and the arp/ndp -an command displays empty L2
tables.

Reviewed by:   pjd

15 years agoEliminate the PMAP_INLINE macro. It isn't really used here. If we
imp [Fri, 16 Jan 2009 08:38:03 +0000 (08:38 +0000)]
Eliminate the PMAP_INLINE macro.  It isn't really used here.  If we
need to bring it back, we can.

15 years agoRemove unused variable.
imp [Fri, 16 Jan 2009 08:30:22 +0000 (08:30 +0000)]
Remove unused variable.
Minor style nits.

15 years agoAdd hardware MAC statistics support. Also added some reserved
yongari [Fri, 16 Jan 2009 08:06:55 +0000 (08:06 +0000)]
Add hardware MAC statistics support. Also added some reserved
statistics register definition. Users can get current MAC
statistics from dev.msk.%d.stats sysctl node(%d is unit number of
a device).

15 years agoadd athrd tool that dumps ath hal regulatory info; note this is soon to
sam [Thu, 15 Jan 2009 23:38:21 +0000 (23:38 +0000)]
add athrd tool that dumps ath hal regulatory info; note this is soon to
obsolete but commit now so it's available for anyone that's interested

15 years agoRevert revision 186833 and try a different strategy to allow this device to
nwhitehorn [Thu, 15 Jan 2009 23:14:59 +0000 (23:14 +0000)]
Revert revision 186833 and try a different strategy to allow this device to
work when the bus attaches its own children. Instead of hardcoding a unit
number and returning BUS_PROBE_NOWILDCARD, which will break multiple iicbus
systems, check in the probe routine whether the device address is 0. Real
I2C devices will never have this address, but devices added with
BUS_ADD_CHILD() will.

Requested by: jhb
Reviewed by: jhb

15 years ago- pmap_track_modified was retired in r178606. Reintroducing it was a mistake.
gonzo [Thu, 15 Jan 2009 23:03:27 +0000 (23:03 +0000)]
- pmap_track_modified was retired in r178606. Reintroducing it was a mistake.

Spotted by: alc@

15 years agoReflect newly added PCI IDs to bce(4).
delphij [Thu, 15 Jan 2009 22:41:48 +0000 (22:41 +0000)]
Reflect newly added PCI IDs to bce(4).

MFC after: 2 months

15 years agoChanges to match "Marketing Description" from Broadcom for HP servers.
delphij [Thu, 15 Jan 2009 22:28:05 +0000 (22:28 +0000)]
Changes to match "Marketing Description" from Broadcom for HP servers.

Submitted by: davidch
MFC after: 2 months

15 years agoDon't cross-reference gem(4) and hme(4); maybe this made some sense
marius [Thu, 15 Jan 2009 22:06:53 +0000 (22:06 +0000)]
Don't cross-reference gem(4) and hme(4); maybe this made some sense
back when these two were the only two NIC drivers working on sparc64
but it no longer does nowadays.

MFC after: 3 days

15 years agoSync the description of jumbo frame support with reality as of r185812;
marius [Thu, 15 Jan 2009 22:00:22 +0000 (22:00 +0000)]
Sync the description of jumbo frame support with reality as of r185812;
it's also supported with BCM5702 (matched by BGE_ASICREV_BCM5703) but
bge(4) currently doesn't with neither BCM5714 nor BCM5780.

Prodded by: scf
MFC after: 3 days

15 years agoChange dstp to be const and remove bogus cast.
rdivacky [Thu, 15 Jan 2009 20:45:59 +0000 (20:45 +0000)]
Change dstp to be const and remove bogus cast.

Submitted by: christoph.mallon at gmx.de
Approved by: kib (mentor)

15 years agoReduce diffs to p4 that were the result of a mismerge on my part.
imp [Thu, 15 Jan 2009 19:57:45 +0000 (19:57 +0000)]
Reduce diffs to p4 that were the result of a mismerge on my part.

15 years agoSilent userland warnings about missing prototypes.
piso [Thu, 15 Jan 2009 19:35:23 +0000 (19:35 +0000)]
Silent userland warnings about missing prototypes.

Submitted by: Roman Divacky <rdivacky@freebsd.org>

15 years agoIntroduce a local variable and use it instead of passed in parameter
rdivacky [Thu, 15 Jan 2009 18:53:52 +0000 (18:53 +0000)]
Introduce a local variable and use it instead of passed in parameter
to get rid of restrict qualifier discarding. This lets libc compile
cleanly in gnu99 mode.

Suggested by: kib, christoph.mallon at gmx.de
Approved by: kib (mentor)

15 years agoMFp4:
gonzo [Thu, 15 Jan 2009 18:31:36 +0000 (18:31 +0000)]
MFp4:

- Add debug output
- Fix pmap_zero_page and related places: use uncached segments and invalidate
    cache after zeroing memory.
- Do not test for modified bit if it's not neccessary
    (merged from mips-juniper p4 branch)
- Some #includes reorganization

15 years agoName the items of xfer array.
takawata [Thu, 15 Jan 2009 16:04:11 +0000 (16:04 +0000)]
Name the items of xfer array.

15 years agoLock the semaphore identifier lock during semaphore initialization to
kib [Thu, 15 Jan 2009 12:15:46 +0000 (12:15 +0000)]
Lock the semaphore identifier lock during semaphore initialization to
guarantee atomicity of the operation for other semaphore consumers.
In particular, this should guard against access to the semaphore with
not done or partially done MAC label assignment.

Reviewed by: rwatson
MFC after: 1 month

15 years agoMFi386: 187144
nyan [Thu, 15 Jan 2009 11:54:31 +0000 (11:54 +0000)]
MFi386: 187144

  Documentation-only change:

  - add a reference to the config(5) manpage;
  - hopefully clarify the format of the 'env FILENAME' directive.

15 years agoMFp4:
imp [Thu, 15 Jan 2009 08:01:50 +0000 (08:01 +0000)]
MFp4:

Remove Maxmem.  It isn't used elsewhere in the system at this point...
realmem is used instead.

15 years agoCall platform_reset() instead of looping forever on reboot.
imp [Thu, 15 Jan 2009 07:51:17 +0000 (07:51 +0000)]
Call platform_reset() instead of looping forever on reboot.
# We likely need to have a default one of these that jumps to the rom boot
# address that's defined in the MIPS ISA.

15 years agoReverse order of dumpsys and cpu_idle_wakeup to reduce diffs to p4.
imp [Thu, 15 Jan 2009 07:48:37 +0000 (07:48 +0000)]
Reverse order of dumpsys and cpu_idle_wakeup to reduce diffs to p4.

15 years agoMFp4:
imp [Thu, 15 Jan 2009 07:45:40 +0000 (07:45 +0000)]
MFp4:

Remove #if'd 0 code.  It is interfering with other diffs.

15 years agofix a ehci's bug that it's occurred when the xfers are aborted under
weongyo [Thu, 15 Jan 2009 07:11:14 +0000 (07:11 +0000)]
fix a ehci's bug that it's occurred when the xfers are aborted under
heavy loads or working.  It looks this bug exists since r158869
so needs to revert a part of the previous.

Reviewed by: imp
Tested by: sam
MFC after: 3 weeks