]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoThis commit was generated by cvs2svn to compensate for changes in r178476,
jb [Fri, 25 Apr 2008 06:56:31 +0000 (06:56 +0000)]
This commit was generated by cvs2svn to compensate for changes in r178476,
which included commits to RCS files with non-trunk default branches.

16 years agoVendor import of the DTrace app and test suite from OpenSolaris.
jb [Fri, 25 Apr 2008 06:56:31 +0000 (06:56 +0000)]
Vendor import of the DTrace app and test suite from OpenSolaris.

16 years ago - Add an integer argument to idle to indicate how likely we are to wake
jeff [Fri, 25 Apr 2008 05:18:50 +0000 (05:18 +0000)]
 - Add an integer argument to idle to indicate how likely we are to wake
   from idle over the next tick.
 - Add a new MD routine, cpu_wake_idle() to wakeup idle threads who are
   suspended in cpu specific states.  This function can fail and cause the
   scheduler to fall back to another mechanism (ipi).
 - Implement support for mwait in cpu_idle() on i386/amd64 machines that
   support it.  mwait is a higher performance way to synchronize cpus
   as compared to hlt & ipis.
 - Allow selecting the idle routine by name via sysctl machdep.idle.  This
   replaces machdep.cpu_idle_hlt.  Only idle routines supported by the
   current machine are permitted.

Sponsored by: Nokia

16 years agoo Disable HMEDEBUG by default.
marius [Thu, 24 Apr 2008 23:12:03 +0000 (23:12 +0000)]
o Disable HMEDEBUG by default.
o Add CTASSERTs ensuring that HME_NRXDESC and HME_NTXDESC are set to
  legal values.
o Use appropriate maxsize, nsegments and maxsegsize parameters when
  creating DMA tags and correct some comments related to them.
o The FreeBSD bus_dmamap_sync(9) supports ored together flags for quite
  some time now so collapse calls accordingly.
o Add missing BUS_DMASYNC_PREREAD when syncing the control DMA maps in
  hme_rint() and hme_start_locked().
o Keep state of the link state and use it to enable or disable the MAC
  in hme_mii_statchg() accordingly as well as to return early from
  hme_start_locked() in case the link is down.
o Introduce a sc_flags and use it to replace individual members like
  sc_pci.
o Add bus_barrier(9) calls to hme_mac_bitflip(), hme_mii_readreg(),
  hme_mii_writereg() and hme_stop() to ensure the respective bit
  has been written before we starting polling on it and for the right
  bits to change.
o Rather just returning in case hme_mac_bitflip() fails and leaving us
  in an undefined state report the problem and move on; chances are
  the requested configuration will become active shortly after.
o Don't call hme_start_locked() in hme_init_locked() unconditionally
  but only after calls to hme_init_locked() when it's appropriate, i.e.
  in hme_watchdog().
o Add a KASSERT which asserts nsegs is valid also to hme_load_txmbuf().
o In hme_load_txmbuf():
  - use a maximum of the newly introduced HME_NTXSEGS segments instead
    of the incorrect HME_NTXQ, which reflects the maximum TX queue
    length, for loading the mbufs and put the DMA segments back onto
    the stack instead of the softc as 16 should be ok there.
  - use the common errno(2) return values instead of homegrown ones,
  - given that hme_load_txmbuf() is allowed to fail resulting in a
    packet drop for quite some time now implement the functionality of
    hme_txcksum() by means of m_pullup(9), which de-obfuscates the code
    and allows to always retrieve the correct length of the IP header, [1]
  - also add a KASSERT which asserts nsegs is valid,
  - take advantage of m_collapse(9) instead of m_defrag(9) for
    performance reasons.
o Don't bother to check whether the interface is running or whether its
  queue is empty before calling hme_start_locked() in hme_tint(), the
  former will check these anyway.
o In hme_intr() call hme_rint() before hme_tint() as gem_tint() may
  take quite a while to return when it calls hme_start_locked().
o Get rid of sc_debug and just check if_flags for IFF_DEBUG directly.
o Add a shadow sc_ifflags so we don't reset the chip when unnecessary.
o Handle IFF_ALLMULTI correctly. [2]
o Use PCIR_BAR instead of a homegrown macro.
o Replace sc_enaddr[6] with sc_enaddr[ETHER_ADDR_LEN].
o Use the maximum of 256 TX descriptors for better performance as using
  all of them has no additional static cost rather than using just half
  of them.

Reported by: rwatson [2]
Suggested by: yongari [1]
Reviewed by: yongari
MFC after: 1 month

16 years ago- vx_stop() requires the driver lock to be held.
marius [Thu, 24 Apr 2008 22:51:43 +0000 (22:51 +0000)]
- vx_stop() requires the driver lock to be held.
- Add some missing newlines to messages printed by vx_getlink().

MFC after: 3 days

16 years ago- Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9)
marius [Thu, 24 Apr 2008 22:48:34 +0000 (22:48 +0000)]
- Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9)
  in order to get rid of bus space handle and tag in struct sym_hcb.
- Remove unused members related to bus addresses in struct sym_hcb.
- sym(4) takes care of allocating an instance of struct sym_hcb
  itself so don't let newbus allocate it as an unused softc also.
- Add basic MPSAFE locking. This includes changing the sym(4) CCBs
  to be allocated up-front instead of on demand as needed. Besides
  making these allocations more likely to succeed, this also solves
  the problem of calling bus_dmamap_create(9) with the SIM mutex
  held.

Reviewed by: scottl
MFC after: 1 month

16 years ago- Use NULL instead of 0 for pointers.
marius [Thu, 24 Apr 2008 22:36:11 +0000 (22:36 +0000)]
- Use NULL instead of 0 for pointers.
- Remove superfluous returns in functions returning void.
- In sym_alloc_lcb_tags() return directly instead of jumping
  to a label which just returns.
- Fix some spelling in comments.
- Remove trailing whitespace.

16 years agofdhold can return NULL, so add the one remaining missing check for this
kris [Thu, 24 Apr 2008 22:08:36 +0000 (22:08 +0000)]
fdhold can return NULL, so add the one remaining missing check for this
condition.

Reviewed by:    attilio
MFC after:      1 week

16 years agoList of files that are off the vendor branch, but we use the stock vendor
obrien [Thu, 24 Apr 2008 18:23:13 +0000 (18:23 +0000)]
List of files that are off the vendor branch, but we use the stock vendor
files.

16 years agoRemove FreeBSD ID's so these are exact copies of the vendor sources.
obrien [Thu, 24 Apr 2008 18:18:53 +0000 (18:18 +0000)]
Remove FreeBSD ID's so these are exact copies of the vendor sources.

16 years agoUse logic or, not binary or, when deciding whether or not a system call
rwatson [Thu, 24 Apr 2008 12:23:31 +0000 (12:23 +0000)]
Use logic or, not binary or, when deciding whether or not a system call
exit requires entering the audit code.  The result is much the same,
but they mean different things.

MFC afer: 3 days
Submitted by: Diego Giagio <dgiagio at gmail dot com>

16 years agoAllow the vnode zone to return the unused memory. The vnode reference
kib [Thu, 24 Apr 2008 09:58:33 +0000 (09:58 +0000)]
Allow the vnode zone to return the unused memory. The vnode reference
count is/shall be properly maintained for the long time, and VFS
shall be safe against the vnode memory reclamation.

Proposed by: jeff
Tested by: pho

16 years agoStricter check for integer overflow.
ru [Thu, 24 Apr 2008 07:49:00 +0000 (07:49 +0000)]
Stricter check for integer overflow.

16 years agoUpdated the test to handle multiple sinks. The sinks cut their
gnn [Thu, 24 Apr 2008 06:56:45 +0000 (06:56 +0000)]
Updated the test to handle multiple sinks.  The sinks cut their
transmissions by the number of them running so that they do not
overwhelm the source.

Added a simple shell script to kick off sinks on multiple hosts as
well as a source on the host where the shell script is run.  The script
also collects the output of all the sinks and the source into files named
for the host on which the tests are run.  A date is appended to each output
file to make it unique per run.

16 years agoIn function rtalloc_mpath(), do not try to release the lock if the ro_rt
qingli [Thu, 24 Apr 2008 05:04:52 +0000 (05:04 +0000)]
In function rtalloc_mpath(), do not try to release the lock if the ro_rt
pointer is NULL.

Reported by: (pluknet at gmail dot com)

16 years agoCapitalize "Eve". This is the correct form and now matches
scf [Thu, 24 Apr 2008 01:37:12 +0000 (01:37 +0000)]
Capitalize "Eve".  This is the correct form and now matches
calendar.usholiday.

MFC after: 3 days

16 years agoAdd support for gpart:
marcel [Thu, 24 Apr 2008 00:11:15 +0000 (00:11 +0000)]
Add support for gpart:
o  Correct for gpart's 1-based index, versus 0-based index used by
   legacy slicers.
o  Parse and understand the xs and xt parameters.

16 years agoRevert rev 1.332 and keep ddb scripts off by default for now. Minidumps
brooks [Wed, 23 Apr 2008 22:40:59 +0000 (22:40 +0000)]
Revert rev 1.332 and keep ddb scripts off by default for now.  Minidumps
are more flexable and much text-dump like output can be produced from
them so there's a good argument they are a better default.

16 years agoAvoid various shadowed variables. libthr is now almost WARNS=4 clean except
delphij [Wed, 23 Apr 2008 21:06:51 +0000 (21:06 +0000)]
Avoid various shadowed variables.  libthr is now almost WARNS=4 clean except
for some const dequalifiers that needs more careful investigation.

Ok'ed by: davidxu

16 years ago- Include <machine/utrap.h> so this header doesn't have an MD
marius [Wed, 23 Apr 2008 20:38:37 +0000 (20:38 +0000)]
- Include <machine/utrap.h> so this header doesn't have an MD
  dependency.
- Make prototypes style(9) compliant.

MFC after: 1 week

16 years agoImplement the G_PART_DUMPCONF method for all 6 schemes. Also call
marcel [Wed, 23 Apr 2008 20:13:05 +0000 (20:13 +0000)]
Implement the G_PART_DUMPCONF method for all 6 schemes. Also call
the method for the (indent == NULL) case (i.e. the kern.geom.conftxt
sysctl). The purpose is to extend the conftxt output with scheme-
specific fields which can be used by libdisk. In particular, have
the schemes dump the xs and xt fields, which contain the backward
compatible values for class type and partition type. This allows
libdisk to work with the legacy slicers as well as with gpart and
helps/promotes migration.

16 years agoo Rename ic_eoi to ic_clear to emphasize the functions it points
marius [Wed, 23 Apr 2008 20:04:38 +0000 (20:04 +0000)]
o Rename ic_eoi to ic_clear to emphasize the functions it points
  don't send and EOI which works like on amd64/i386 and blocks all
  interrupts on the relevant interrupt controller.
o Replace the post_filter and post_inthread hooks registered when
  creating the interrupt events with just ic_clear as on sparc64 we
  don't need to do any disable->EOI->enable dance to unblock all but
  the relevant interrupt while running the filter or handler; just
  not clearing the interrupt already has the same effect.
o Merge from amd64/i386:
  - Split the intr_table_lock into an sx lock used for most things,
    and a spin lock to protect intrcnt_index.
  - Add support for binding interrupts to CPUs, including for the
    bus_bind_intr(9) interface, a assign_cpu hook and initially
    shuffling interrupts arround in a round-robin fashion.

Reviewed by: jhb
MFC after: 1 month

16 years agoImplement red-black trees without using parent pointers, and store the
jasone [Wed, 23 Apr 2008 16:09:18 +0000 (16:09 +0000)]
Implement red-black trees without using parent pointers, and store the
color bit in the least significant bit of the right child pointer, in
order to reduce red-black tree linkage overhead by ~2X as compared to
sys/tree.h.

Use the new red-black tree implementation in malloc, which drops
memory usage by ~0.5 or ~1%, for 32- and 64-bit systems, respectively.

16 years agoImplement linux_truncate64() syscall.
rdivacky [Wed, 23 Apr 2008 15:56:33 +0000 (15:56 +0000)]
Implement linux_truncate64() syscall.

Tested by: Aline de Freitas <aline@riseup.net>
Approved by: kib (mentor)

16 years agoFix the year of my doc/www commit bit that I've got :)
ganbold [Wed, 23 Apr 2008 08:51:05 +0000 (08:51 +0000)]
Fix the year of my doc/www commit bit that I've got :)

Approved by: keramida(mentor)

16 years agoAdd 'arm' category.
linimon [Wed, 23 Apr 2008 03:05:31 +0000 (03:05 +0000)]
Add 'arm' category.

Approved by: peter

16 years agoCorrect an off-by-1 for GPART. The literal partition type (i.e.
marcel [Wed, 23 Apr 2008 03:00:26 +0000 (03:00 +0000)]
Correct an off-by-1 for GPART. The literal partition type (i.e.
the actual UUID) is prefixed by '!' to distinguish them from
well-known aliases.

MFC after: 3 days

16 years agoAdd four utility functions related to struct grp processing modeled in-part
scf [Wed, 23 Apr 2008 00:49:13 +0000 (00:49 +0000)]
Add four utility functions related to struct grp processing modeled in-part
after similar calls related to struct pwd in libutil/pw_util.c:
  - gr_equal()
    Perform a deep comparison of two struct grp's.  It does a thorough, yet
    unoptimized comparison of all the members regardless of order.

  - gr_make()
    Create a string (see group(5)) from a struct grp.

  - gr_dup()
    Duplicate a struct grp.  Returns a value that is a single contiguous
    block of memory.

  - gr_scan()
    Create a struct grp from a string (as produced by gr_make()).

MFC after: 3 weeks

16 years agoDifferentiate in the tests against getenv() between an empty string and a
scf [Wed, 23 Apr 2008 00:45:48 +0000 (00:45 +0000)]
Differentiate in the tests against getenv() between an empty string and a
NULL by having the test program return *NULL* for NULL's (no existing
variable) instead of an empty string (variable with empty value).

16 years agoNow that all platforms use genclock, shuffle things around slightly
phk [Tue, 22 Apr 2008 19:38:30 +0000 (19:38 +0000)]
Now that all platforms use genclock, shuffle things around slightly
for better structure.

Much of this is related to <sys/clock.h>, which should really have
been called <sys/calendar.h>, but unless and until we need the name,
the repocopy can wait.

In general the kernel does not know about minutes, hours, days,
timezones, daylight savings time, leap-years and such.  All that
is theoretically a matter for userland only.

Parts of kernel code does however care: badly designed filesystems
store timestamps in local time and RTC chips almost universally
track time in a YY-MM-DD HH:MM:SS format, and sometimes in local
timezone instead of UTC.  For this we have <sys/clock.h>

<sys/time.h> on the other hand, deals with time_t, timeval, timespec
and so on.  These know only seconds and fractions thereof.

Move inittodr() and resettodr() prototypes to <sys/time.h>.
Retain the names as it is one of the few surviving PDP/VAX references.

Move startrtclock() to <machine/clock.h> on relevant platforms, it
is a MD call between machdep.c/clock.c.  Remove references to it
elsewhere.

Remove a lot of unnecessary <sys/clock.h> includes.

Move the machdep.disable_rtc_set sysctl to subr_rtc.c where it belongs.
XXX: should be kern.disable_rtc_set really, it's not MD.

16 years agolibbsdxml is now required by ifconfig(8).
thompsa [Tue, 22 Apr 2008 18:20:05 +0000 (18:20 +0000)]
libbsdxml is now required by ifconfig(8).

Submitted by: Florian Smeets

16 years agoFix a leak in the recent fixes for file descriptors > SHRT_MAX. In the
jhb [Tue, 22 Apr 2008 17:03:32 +0000 (17:03 +0000)]
Fix a leak in the recent fixes for file descriptors > SHRT_MAX.  In the
case of a file descriptor we can't handle, clear the FILE structure's flags
so it can be reused.

MFC after: 1 week
Reported by: otto @ OpenBSD

16 years agoAdd:
pgj [Tue, 22 Apr 2008 15:42:20 +0000 (15:42 +0000)]
Add:
- myself to the doc committers' graph
- my birthday to the FreeBSD calendar

Approved by: gabor (mentor)

16 years agoMake it possible to disable sources of entropy harvesting.
ru [Tue, 22 Apr 2008 15:18:47 +0000 (15:18 +0000)]
Make it possible to disable sources of entropy harvesting.

Noticed by: Igor Sysoev
MFC after: 3 days

16 years ago - Use a local variable for i_ino in ufs_lookup. It is only used to
jeff [Tue, 22 Apr 2008 12:34:16 +0000 (12:34 +0000)]
 - Use a local variable for i_ino in ufs_lookup.  It is only used to
   communicate between two parts of this one function.  This was causing
   problems with shared lookups as each would trash the ino value in the
   inode.
 - Remove the unused i_ino field from the inode structure.

16 years agoAcquire a read lock, rather than a write lock, on a UDPv6 inpcb when
rwatson [Tue, 22 Apr 2008 12:20:33 +0000 (12:20 +0000)]
Acquire a read lock, rather than a write lock, on a UDPv6 inpcb when
delivering to the socket or extracting socket details for monitoring
purposes.

MFC after: 3 months

16 years agoAdd FreeBSD IDs to files that originate in FreeBSD.
jb [Tue, 22 Apr 2008 07:43:00 +0000 (07:43 +0000)]
Add FreeBSD IDs to files that originate in FreeBSD.

16 years agoUse native rwlock.
davidxu [Tue, 22 Apr 2008 06:44:11 +0000 (06:44 +0000)]
Use native rwlock.

16 years agoKill $P4$ id.
rpaulo [Mon, 21 Apr 2008 22:00:01 +0000 (22:00 +0000)]
Kill $P4$ id.

Noticed by: rwatson

16 years agoThe vmspace->vm_daddr is constant until freed, there is no need
rdivacky [Mon, 21 Apr 2008 21:24:08 +0000 (21:24 +0000)]
The vmspace->vm_daddr is constant until freed, there is no need
to hold lock while accessing it.

Approved by: kib (mentor)

16 years agoThis commit was generated by cvs2svn to compensate for changes in r178388,
marius [Mon, 21 Apr 2008 20:40:42 +0000 (20:40 +0000)]
This commit was generated by cvs2svn to compensate for changes in r178388,
which included commits to RCS files with non-trunk default branches.

16 years ago* gthr-posix.h (__gthread_active_p): Use the Solaris implementation
marius [Mon, 21 Apr 2008 20:40:42 +0000 (20:40 +0000)]
* gthr-posix.h (__gthread_active_p): Use the Solaris implementation
for FreeBSD as well.

This is the fix for __gthread_active_p() returning false positives
which was committed as rev. 1.1.1.8.2.1 to RELENG_7 but now looped
back to the vendor branch via the GCC repository and relicensed to
be GPLv2 by me.
Thanks go to gerald@ for getting the fix approved upstream and for
committing to the GCC repository.

PR: 119289
Approved by: core

16 years agoUpdate instructions to match recent practices
delphij [Mon, 21 Apr 2008 18:37:08 +0000 (18:37 +0000)]
Update instructions to match recent practices

16 years agoAdd vendor metadata
delphij [Mon, 21 Apr 2008 18:36:52 +0000 (18:36 +0000)]
Add vendor metadata

16 years agoResolve conflicts.
delphij [Mon, 21 Apr 2008 18:31:50 +0000 (18:31 +0000)]
Resolve conflicts.

16 years agoThis commit was generated by cvs2svn to compensate for changes in r178382,
delphij [Mon, 21 Apr 2008 18:30:26 +0000 (18:30 +0000)]
This commit was generated by cvs2svn to compensate for changes in r178382,
which included commits to RCS files with non-trunk default branches.

16 years agoImport netcat as of today's OPENBSD_4_3 snapshot.
delphij [Mon, 21 Apr 2008 18:30:26 +0000 (18:30 +0000)]
Import netcat as of today's OPENBSD_4_3 snapshot.

16 years agoChange the default of ddb_enable to YES so we default to generating textdumps
brooks [Mon, 21 Apr 2008 18:17:48 +0000 (18:17 +0000)]
Change the default of ddb_enable to YES so we default to generating textdumps
on panic.  This means you get a potentially useful dump even if your system
is running X when you panic.

X-MFC after: never

16 years agoBack-out previous revision. For now I can use _ddb() variants of stack(9) KPI,
pjd [Mon, 21 Apr 2008 17:22:35 +0000 (17:22 +0000)]
Back-out previous revision. For now I can use _ddb() variants of stack(9) KPI,
as I use it for debugging only. Once someone will need it for more production
features, the change should be reconsider.

Requested by: rwatson

16 years agoUse ddb(4), not DDB(4) for man page cross-references.
rwatson [Mon, 21 Apr 2008 17:09:53 +0000 (17:09 +0000)]
Use ddb(4), not DDB(4) for man page cross-references.

MFC after: 3 days
Reported by: novel

16 years agoIn ICMPv6, read lock rather than write lock the inpcb on receive.
rwatson [Mon, 21 Apr 2008 12:08:40 +0000 (12:08 +0000)]
In ICMPv6, read lock rather than write lock the inpcb on receive.

MFC after: 3 months

16 years agoWith IPv4 raw sockets, read lock rather than write lock the inpcb when
rwatson [Mon, 21 Apr 2008 12:06:41 +0000 (12:06 +0000)]
With IPv4 raw sockets, read lock rather than write lock the inpcb when
receiving or transmitting.

With IPv6 raw sockets, read lock rather than write lock the inpcb when
receiving.  Unfortunately, IPv6 source address selection appears to
require a write lock on the inpcb for the time being.

MFC after: 3 months

16 years agoRead lock, rather than write lock, the inpcb when transmitting with or
rwatson [Mon, 21 Apr 2008 12:03:59 +0000 (12:03 +0000)]
Read lock, rather than write lock, the inpcb when transmitting with or
delivering to an IP divert socket.

MFC after: 3 months

16 years agoAdd HW level support for the Adaptec 1420SA controller.
sos [Mon, 21 Apr 2008 10:51:38 +0000 (10:51 +0000)]
Add HW level support for the Adaptec 1420SA controller.

16 years agoUnload DMA slot on device timeouts.
sos [Mon, 21 Apr 2008 10:35:19 +0000 (10:35 +0000)]
Unload DMA slot on device timeouts.

16 years agoDo not enable FIS based PM switching, even if not used some controllers barf.
sos [Mon, 21 Apr 2008 10:34:31 +0000 (10:34 +0000)]
Do not enable FIS based PM switching, even if not used some controllers barf.

16 years agoMake genclock standard on all platforms.
phk [Mon, 21 Apr 2008 10:09:55 +0000 (10:09 +0000)]
Make genclock standard on all platforms.

Thanks to: grehan & marcel for platform support on ia64 and ppc.

16 years agoRemove duplicate headers <sys/socket.h>
kevlo [Mon, 21 Apr 2008 07:25:26 +0000 (07:25 +0000)]
Remove duplicate headers <sys/socket.h>

16 years agoAdd a small script that will examine all files which exist on /cfg and
phk [Mon, 21 Apr 2008 06:22:27 +0000 (06:22 +0000)]
Add a small script that will examine all files which exist on /cfg and
if they are different in /etc, copy them to /cfg.

Submitted by: mat

16 years agoawi(4) has been removed.
brueffer [Mon, 21 Apr 2008 05:43:08 +0000 (05:43 +0000)]
awi(4) has been removed.

16 years agoSwitch to using genclock. Have nexus double as clock device for
marcel [Mon, 21 Apr 2008 04:41:37 +0000 (04:41 +0000)]
Switch to using genclock. Have nexus double as clock device for
now. While here, add a proper attach() method to nexus.

Requested by: phk

16 years agoOn the AT91, we need to write on the EOI register after we handle an
cognet [Sun, 20 Apr 2008 23:29:06 +0000 (23:29 +0000)]
On the AT91, we need to write on the EOI register after we handle an
interrupt. So, add a new function pointer, arm_post_filter, which defaults
to NULL, and which will be used as the post_filter arg for
intr_event_create(). Set it properly for the AT91, so that it boots again.

Reported by: hps

16 years agofix build
sam [Sun, 20 Apr 2008 23:10:44 +0000 (23:10 +0000)]
fix build

16 years agoThis commit was generated by cvs2svn to compensate for changes in r178363,
sam [Sun, 20 Apr 2008 21:39:06 +0000 (21:39 +0000)]
This commit was generated by cvs2svn to compensate for changes in r178363,
which included commits to RCS files with non-trunk default branches.

16 years ago802.11 support moves to vaps
sam [Sun, 20 Apr 2008 21:25:37 +0000 (21:25 +0000)]
802.11 support moves to vaps

16 years agotest scripts for vaps and other 802.11 features
sam [Sun, 20 Apr 2008 20:43:42 +0000 (20:43 +0000)]
test scripts for vaps and other 802.11 features

Supported by: Hobnob and Marvell

16 years agoadd dwds test app
sam [Sun, 20 Apr 2008 20:43:13 +0000 (20:43 +0000)]
add dwds test app

Supported by: Hobnob
Obtained from: Atheros (earlier version)

16 years agoupdate for vaps
sam [Sun, 20 Apr 2008 20:41:47 +0000 (20:41 +0000)]
update for vaps

Supported by: Hobnob

16 years agoo update for vaps
sam [Sun, 20 Apr 2008 20:40:45 +0000 (20:40 +0000)]
o update for vaps
o add private wired driver that fixes various bugs in the vendor version

Submitted by: thompsa (ndis fixups)

16 years agoo update for vaps
sam [Sun, 20 Apr 2008 20:39:08 +0000 (20:39 +0000)]
o update for vaps
o add+enable radius acl support

Supported by: Hobnob
Submitted by: Chris Zimmermann (acl support)

16 years agorc support for vaps
sam [Sun, 20 Apr 2008 20:37:21 +0000 (20:37 +0000)]
rc support for vaps

16 years ago802.11 regulatory definitions used by ifconfig
sam [Sun, 20 Apr 2008 20:37:02 +0000 (20:37 +0000)]
802.11 regulatory definitions used by ifconfig

Support by: Hobnob

16 years agoMulti-bss (aka vap) support for 802.11 devices.
sam [Sun, 20 Apr 2008 20:35:46 +0000 (20:35 +0000)]
Multi-bss (aka vap) support for 802.11 devices.

Note this includes changes to all drivers and moves some device firmware
loading to use firmware(9) and a separate module (e.g. ral).  Also there
no longer are separate wlan_scan* modules; this functionality is now
bundled into the wlan module.

Supported by: Hobnob and Marvell
Reviewed by: many
Obtained from: Atheros (some bits)

16 years agomove awi to the Attic; it will not make the jump to the new world order
sam [Sun, 20 Apr 2008 19:20:39 +0000 (19:20 +0000)]
move awi to the Attic; it will not make the jump to the new world order

Reviewed by: imp

16 years ago- Add ($paragraph$) for printed version.
hrs [Sun, 20 Apr 2008 18:41:16 +0000 (18:41 +0000)]
- Add ($paragraph$) for printed version.
- Fix a bug that role="" does not work when arch="" is not specified.

16 years agoRevert to rev. 1.161 - switch back to optimized TCP options ordering.
bz [Sun, 20 Apr 2008 18:36:59 +0000 (18:36 +0000)]
Revert to rev. 1.161 - switch back to optimized TCP options ordering.

A lot of testing has shown that the problem people were seeing was due
to invalid padding after the end of option list option, which was corrected
in tcp_output.c rev. 1.146.

Thanks to: anders@, s3raphi, Matt Reimer
Thanks to: Doug Hardie and Randy Rose, John Mayer, Susan Guzzardi
Special thanks to: dwhite@ and BitGravity
Discussed with: silby
MFC after: 1 day

16 years agoMove (define ($paragraph$)) to the marked section for HTML output.
hrs [Sun, 20 Apr 2008 18:28:31 +0000 (18:28 +0000)]
Move (define ($paragraph$)) to the marked section for HTML output.
It should not be applied to printed output.

16 years agoNew release notes:
hrs [Sun, 20 Apr 2008 18:17:33 +0000 (18:17 +0000)]
New release notes:
ULE scheduler enabled by default in GENERIC for ia64,
fdopendir(3) added, and
tcgetsid(3) added.

16 years agoNew release notes:
hrs [Sun, 20 Apr 2008 18:03:21 +0000 (18:03 +0000)]
New release notes:
env(1) -u name option.

16 years agoUse role="merged" and arch="amd64,i386,..." attributes to mark a
hrs [Sun, 20 Apr 2008 17:58:05 +0000 (17:58 +0000)]
Use role="merged" and arch="amd64,i386,..." attributes to mark a
paragraph as MFC'd or one for specific platforms.  This should
make no content change with the rendered output.

16 years agodevclass_get_maxunit() returns n+1 with n starting at 0.
bz [Sun, 20 Apr 2008 17:45:32 +0000 (17:45 +0000)]
devclass_get_maxunit() returns n+1 with n starting at 0.
So if we have channel 0..3  devclass_get_maxunit is 4.

It's never been a problem as devclass_get_device() has
catched a possibly bad input.

Discussed with: scottl

16 years agoAdd one more HTC PocketPC device id.
mav [Sun, 20 Apr 2008 17:12:11 +0000 (17:12 +0000)]
Add one more HTC PocketPC device id.
Tested on HTC TyTN.

Submitted by: Alexander Logvinov <ports@logvinov.com>

16 years agoMove "1000baseT" from IFM_SUBTYPE_ETHERNET_DESCRIPTIONS to
antoine [Sun, 20 Apr 2008 16:17:44 +0000 (16:17 +0000)]
Move "1000baseT" from IFM_SUBTYPE_ETHERNET_DESCRIPTIONS to
IFM_SUBTYPE_ETHERNET_ALIASES: there is already "1000baseTX" in
IFM_SUBTYPE_ETHERNET_DESCRIPTIONS.  This doesn't change ifconfig
behaviour.

PR: 45793 (maybe)
Approved by: rwatson (mentor)
MFC after: 1 month

16 years agoUnbreak firmware.ko build without kernel.
antoine [Sun, 20 Apr 2008 16:11:14 +0000 (16:11 +0000)]
Unbreak firmware.ko build without kernel.

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

16 years agoAdd recent obsolete files.
antoine [Sun, 20 Apr 2008 16:03:19 +0000 (16:03 +0000)]
Add recent obsolete files.

Approved by: rwatson (mentor)

16 years agoRecord the dependency of the iscsi initiator on cam.
antoine [Sun, 20 Apr 2008 15:56:59 +0000 (15:56 +0000)]
Record the dependency of the iscsi initiator on cam.

PR: 116745
Submitted by: OISHI Masakuni
Approved by: rwatson (mentor)
MFC after: 1 month

16 years agoDocument that you must include <sys/param.h> before <sys/cpuset.h>.
antoine [Sun, 20 Apr 2008 15:51:56 +0000 (15:51 +0000)]
Document that you must include <sys/param.h> before <sys/cpuset.h>.

Approved by: rwatson (mentor)

16 years agoTeach pf and ipfw to use read locks in inpcbs write than write locks
rwatson [Sun, 20 Apr 2008 00:21:54 +0000 (00:21 +0000)]
Teach pf and ipfw to use read locks in inpcbs write than write locks
when reading credential data from sockets.

Teach pf to unlock the pcbinfo more quickly once it has acquired an
inpcb lock, as the inpcb lock is sufficient to protect the reference.

Assert locks, rather than read locks or write locks, on inpcbs in
subroutines--this is necessary as the inpcb may be passed down with a
write lock from the protocol, or may be passed down with a read lock
from the firewall lookup routine, and either is sufficient.

MFC after: 3 months

16 years agoFix no sound issues for ASUS W6F (ALC861) laptops (buggy BIOS config that
ariff [Sat, 19 Apr 2008 22:32:23 +0000 (22:32 +0000)]
Fix no sound issues for ASUS W6F (ALC861) laptops (buggy BIOS config that
deserves its own internet memes). The trick is to force all available,
unused pins (that being advertised as "speaker") to behave as microphone
pins instead.

Reported / Tested by: Dmitry Kutsenko <kutsenko.truebsd.org>
MFC after: 3 days

16 years agoDelay the global registration of the struct ifnet in if_alloc() until after
brooks [Sat, 19 Apr 2008 22:04:51 +0000 (22:04 +0000)]
Delay the global registration of the struct ifnet in if_alloc() until after
we're certain the allocation will entierly succeed.  This fixes a leak in a
fairly unlikely case.

Reported by: vijay singh <vijjus at rocketmail dot com>
MFC after: 1 week

16 years agoWhen auditing state from an IPv4 or IPv6 socket, use read locks on the
rwatson [Sat, 19 Apr 2008 18:37:08 +0000 (18:37 +0000)]
When auditing state from an IPv4 or IPv6 socket, use read locks on the
inpcb rather than write locks.

MFC after: 3 months

16 years agoWhen propagating a MAC label from an inpcb to an mbuf, allow read and
rwatson [Sat, 19 Apr 2008 18:35:27 +0000 (18:35 +0000)]
When propagating a MAC label from an inpcb to an mbuf, allow read and
write locks on the inpcb, not just write locks.

MFC after: 3 months

16 years agoWhen querying a local or remote address on an IPv6 socket, use only a
rwatson [Sat, 19 Apr 2008 14:36:19 +0000 (14:36 +0000)]
When querying a local or remote address on an IPv6 socket, use only a
read lock on the inpcb.

MFC after: 3 months

16 years agoIn ip_output(), allow a read lock as well as a write lock when asserting
rwatson [Sat, 19 Apr 2008 14:35:17 +0000 (14:35 +0000)]
In ip_output(), allow a read lock as well as a write lock when asserting
a lock on the passed inpcb.

MFC after: 3 months

16 years agoWhen querying the local or foreign address from an IP socket, acquire
rwatson [Sat, 19 Apr 2008 14:34:38 +0000 (14:34 +0000)]
When querying the local or foreign address from an IP socket, acquire
only a read lock on the inpcb.

When an external module requests a read lock, acquire only a read lock.

MFC after: 3 months

16 years agoProvide more detailed information about each procstat(1) display mode,
rwatson [Sat, 19 Apr 2008 13:40:42 +0000 (13:40 +0000)]
Provide more detailed information about each procstat(1) display mode,
including a key to fields in each mode and flag abbreviations.

MFC after: 3 days
X-MFC-note: POSIX shared memory memory objects aren't in 7-STABLE yet

16 years agoIt is a bug that procstat(8) works only on live kernels and not crashdumps;
rwatson [Sat, 19 Apr 2008 12:39:15 +0000 (12:39 +0000)]
It is a bug that procstat(8) works only on live kernels and not crashdumps;
document in case anyone wants to work on fixing this.

MFC after: 3 days

16 years agoMFi386: Merge yet another the RTC related work.
nyan [Sat, 19 Apr 2008 08:18:47 +0000 (08:18 +0000)]
MFi386: Merge yet another the RTC related work.

Split the pcrtc driver into pcrtc.c which is repo-copied from clock.c

16 years agoPut in a real isa_irq_pending() stub in order to remove two lines of dmesg
peter [Sat, 19 Apr 2008 07:25:57 +0000 (07:25 +0000)]
Put in a real isa_irq_pending() stub in order to remove two lines of dmesg
noise from sio per unit.  sio likes to probe if interrupts are configured
correctly by looking at the pending bits of the atpic in order to put a
non-fatal warning on the console.  I think I'd rather read the pending
bits from the apics, but I'm not sure its worth the hassle.

16 years agoDon't forget to free() currency_symbol and asciivalue when multiple
ru [Sat, 19 Apr 2008 07:22:58 +0000 (07:22 +0000)]
Don't forget to free() currency_symbol and asciivalue when multiple
conversion specifiers for them are present.

Submitted by: Maxim Dounin <mdounin@mdounin.ru>
Obtained from: NetBSD (partially)
MFC after: 3 days