]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
13 years agoAdd stablerestart(5) to the See Also list for nfsd(8).
rmacklem [Tue, 19 Apr 2011 00:47:26 +0000 (00:47 +0000)]
Add stablerestart(5) to the See Also list for nfsd(8).
This is a content change.

Suggested by: Jeremy Chadwick
MFC after: 2 weeks

13 years agoFurthermore condition IP waiting behaviour also on -dynamic flag.
sobomax [Mon, 18 Apr 2011 23:45:50 +0000 (23:45 +0000)]
Furthermore condition IP waiting behaviour also on -dynamic flag.
This should preserve POLA in all other cases (foreground || non-dynamic).

MFC after: 2 weeks

13 years agoRevert r220761 since, as kib@ pointed out, the case of
rmacklem [Mon, 18 Apr 2011 23:35:16 +0000 (23:35 +0000)]
Revert r220761 since, as kib@ pointed out, the case of
adding the check to nfsrpc_close() isn't useful. Also,
the check in nfscl_getcl() must be more involved, since
it needs to check before and after the acquisition of
the refcnt on nfsc_lock, while the mutex that protects
the client state data is held.

13 years agoOnly wait for the IP to appear on the interface if natd is running in the
sobomax [Mon, 18 Apr 2011 23:15:29 +0000 (23:15 +0000)]
Only wait for the IP to appear on the interface if natd is running in the
background.

Suggested by:   Garrett Cooper <yanegomi@gmail.com>

Use EAGAIN instead of magic value of -2 to report this condition from the
SetAliasAddressFromIfName routine.

MFC after: 2 weeks

13 years agoDo not assume PM timer GAS type is I/O or memory. It may be an unsupported
jkim [Mon, 18 Apr 2011 23:12:41 +0000 (23:12 +0000)]
Do not assume PM timer GAS type is I/O or memory.  It may be an unsupported
type, i. e., a broken table.  Also, do not hardcode ACPI timer frequency in
device description.

13 years agoMore whitespace fixes.
glebius [Mon, 18 Apr 2011 22:09:03 +0000 (22:09 +0000)]
More whitespace fixes.

Checked with: md5, diff -x -w

13 years agoMake pmap_invalidate_cache_range() available for consumption on amd64.
kib [Mon, 18 Apr 2011 21:24:42 +0000 (21:24 +0000)]
Make pmap_invalidate_cache_range() available for consumption on amd64.

Add pmap_invalidate_cache_pages() method on x86. It flushes the CPU
cache for the set of pages, which are not neccessary mapped. Since its
supposed use is to prepare the move of the pages ownership to a device
that does not snoop all CPU accesses to the main memory (read GPU in
GMCH), do not rely on CPU self-snoop feature.

amd64 implementation takes advantage of the direct map. On i386,
extract the helper pmap_flush_page() from pmap_page_set_memattr(), and
use it to make a temporary mapping of the flushed page.

Reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 3 weeks

13 years agoWhitespace fixes.
glebius [Mon, 18 Apr 2011 21:18:22 +0000 (21:18 +0000)]
Whitespace fixes.

Checked with: md5, diff -w

13 years agoChanged "conscontrol unset" to accept an existing virtual
ru [Mon, 18 Apr 2011 20:28:07 +0000 (20:28 +0000)]
Changed "conscontrol unset" to accept an existing virtual
console device as an argument.  Unsetting virtual console
using /dev/console seems to have never worked.

MFC after: 3 days

13 years agoLibAliasInit() should allocate memory with M_WAITOK flag. Modify it
glebius [Mon, 18 Apr 2011 20:07:08 +0000 (20:07 +0000)]
LibAliasInit() should allocate memory with M_WAITOK flag. Modify it
and its callers.

13 years agoDocument problems with -d/-w and the fact that -X is the default.
trasz [Mon, 18 Apr 2011 19:20:47 +0000 (19:20 +0000)]
Document problems with -d/-w and the fact that -X is the default.

Suggested by: arundel@
Reviewed by: arundel@

13 years agoFix a few acpi sysctls that want "IK" formatting to specify CTLTYPE_INT.
mdf [Mon, 18 Apr 2011 19:02:41 +0000 (19:02 +0000)]
Fix a few acpi sysctls that want "IK" formatting to specify CTLTYPE_INT.
This got broken after r217586.

Pointy hat: to me
Tested by: David Wolfskill < davit AT catwhisker DOT org >

13 years agoPullup up to TCP header length before matching against 'tcpopts'.
glebius [Mon, 18 Apr 2011 18:22:10 +0000 (18:22 +0000)]
Pullup up to TCP header length before matching against 'tcpopts'.

PR: kern/156180
Reviewed by: luigi

13 years agoWhen checking to see if a window update should be sent to the remote peer,
jhb [Mon, 18 Apr 2011 17:43:16 +0000 (17:43 +0000)]
When checking to see if a window update should be sent to the remote peer,
don't force a window update if the window would not actually grow due to
window scaling.  Specifically, if the window scaling factor is larger than
2 * MSS, then after the local reader has drained 2 * MSS bytes from the
socket, a window update can end up advertising the same window.  If this
happens, the supposed window update actually ends up being a duplicate ACK.
This can result in an excessive number of duplicate ACKs when using a
higher maximum socket buffer size.

Reviewed by: bz
MFC after: 1 month

13 years agoFix a copy/paste whitespace error.
mdf [Mon, 18 Apr 2011 16:40:47 +0000 (16:40 +0000)]
Fix a copy/paste whitespace error.

13 years agoRegen.
mdf [Mon, 18 Apr 2011 16:32:47 +0000 (16:32 +0000)]
Regen.

13 years agoAdd the posix_fallocate(2) syscall. The default implementation in
mdf [Mon, 18 Apr 2011 16:32:22 +0000 (16:32 +0000)]
Add the posix_fallocate(2) syscall.  The default implementation in
vop_stdallocate() is filesystem agnostic and will run as slow as a
read/write loop in userspace; however, it serves to correctly
implement the functionality for filesystems that do not implement a
VOP_ALLOCATE.

Note that __FreeBSD_version was already bumped today to 900036 for any
ports which would like to use this function.

Also reserve space in the syscall table for posix_fadvise(2).

Reviewed by: -arch (previous version)

13 years agoReduce geom_raid log verbosity.
mav [Mon, 18 Apr 2011 16:15:59 +0000 (16:15 +0000)]
Reduce geom_raid log verbosity.

13 years agoHandle ready timeout during polled operation same as done in mvs(4) before.
mav [Mon, 18 Apr 2011 16:00:46 +0000 (16:00 +0000)]
Handle ready timeout during polled operation same as done in mvs(4) before.

13 years agoIt is already seven years since mbuf allocator uses same
glebius [Mon, 18 Apr 2011 15:30:25 +0000 (15:30 +0000)]
It is already seven years since mbuf allocator uses same
M_WAITOK/M_NOWAIT flags as malloc(9). Update manual page.

Submitted by: Vadim Goncharov <vadimnuclight tpu.ru>

13 years agoRemove always false "< 0" check for unsgined int variable. This check is
mav [Mon, 18 Apr 2011 14:34:10 +0000 (14:34 +0000)]
Remove always false "< 0" check for unsgined int variable. This check is
also duplicate, as the value was already checked for 0 before decrementing.

Reported by: rpaulo

13 years agoRemove some used variables.
mav [Mon, 18 Apr 2011 14:15:52 +0000 (14:15 +0000)]
Remove some used variables.

Found with: Clang Static Analyzer

13 years agoFor now, only enable GTT. CST is firing very frequently during local tests;
adrian [Mon, 18 Apr 2011 14:14:54 +0000 (14:14 +0000)]
For now, only enable GTT. CST is firing very frequently during local tests;
I'll figure out what's going on before re-enabling this as it does add
to the interrupt load.

13 years agoFinish last change.
glebius [Mon, 18 Apr 2011 14:07:01 +0000 (14:07 +0000)]
Finish last change.

Pointy hat to: glebius

13 years agoAdd TX carrier sense timeout statistics.
adrian [Mon, 18 Apr 2011 14:06:18 +0000 (14:06 +0000)]
Add TX carrier sense timeout statistics.

13 years agoFurther cleanup of node creation path from M_NOWAIT usage.
glebius [Mon, 18 Apr 2011 14:05:26 +0000 (14:05 +0000)]
Further cleanup of node creation path from M_NOWAIT usage.

13 years agoBump pad, I'm adding more statistics.
adrian [Mon, 18 Apr 2011 14:03:37 +0000 (14:03 +0000)]
Bump pad, I'm adding more statistics.

13 years agoRework the Global TX timeout handling to look more like ath9k.
adrian [Mon, 18 Apr 2011 14:03:05 +0000 (14:03 +0000)]
Rework the Global TX timeout handling to look more like ath9k.

It correctly now sets the AR_IMR BCNMISC register, along with
the GTT register in AR_IMR_S2.

13 years agoRemove some used variables.
mav [Mon, 18 Apr 2011 13:59:56 +0000 (13:59 +0000)]
Remove some used variables.

Found with: Clang Static Analyzer

13 years ago - Tune different wait loops to cut some more milliseconds from reset time.
mav [Mon, 18 Apr 2011 13:34:31 +0000 (13:34 +0000)]
 - Tune different wait loops to cut some more milliseconds from reset time.
 - Do not call ahci_start() before device signature received. It is required
by the specification and caused non-fatal reset timeouts on AMD chipsets.

13 years agoMention the memstick target here.
nwhitehorn [Mon, 18 Apr 2011 13:20:54 +0000 (13:20 +0000)]
Mention the memstick target here.

13 years agoDelete mistakenly added sys/files.ts7800
philip [Mon, 18 Apr 2011 12:27:57 +0000 (12:27 +0000)]
Delete mistakenly added sys/files.ts7800
Add mistakenly forgotten sys/arm/conf/TS7800

Not sure how this happened.  Apologies for the repo-churn.

Submitted by: glebius
Pointy hat to: philip

13 years agoAdd global TX timeout handling.
adrian [Mon, 18 Apr 2011 12:15:43 +0000 (12:15 +0000)]
Add global TX timeout handling.

The global TX timeout counter increments whenever a frame is ready
to be transmitted and the medium is busy.

13 years agoAfter removing libobjc and other Objective-C components in r220755,
dim [Mon, 18 Apr 2011 10:25:54 +0000 (10:25 +0000)]
After removing libobjc and other Objective-C components in r220755,
belatedly bump __FreeBSD_version, and add a note to UPDATING.

Reminded by: rdivacky

13 years agong_netflow_cache_init() can be void.
glebius [Mon, 18 Apr 2011 09:14:23 +0000 (09:14 +0000)]
ng_netflow_cache_init() can be void.

13 years agoNode constructor methods are supposed to be called in syscall
glebius [Mon, 18 Apr 2011 09:12:27 +0000 (09:12 +0000)]
Node constructor methods are supposed to be called in syscall
context always. Convert nodes to consistently use M_WAITOK flag
for memory allocation.

Reviewed by: julian

13 years agoUse M_WAITOK flag instead M_WAIT for malloc.
ae [Mon, 18 Apr 2011 09:10:27 +0000 (09:10 +0000)]
Use M_WAITOK flag instead M_WAIT for malloc.

Suggested by: glebius
MFC after: 1 week

13 years agoAdd a vput() to nfs_lookitup() in the experimental NFS client
rmacklem [Mon, 18 Apr 2011 01:02:43 +0000 (01:02 +0000)]
Add a vput() to nfs_lookitup() in the experimental NFS client
for a case that will probably never happen. It can only
happen if a server were to successfully lookup a file, but not
return attributes for that file. Although technically allowed
by the NFSv3 RFC, I doubt any server would ever do this.
However, if it did, the client would have not vput()'d the
new vnode when it needed to do so.

MFC after: 2 weeks

13 years agoAdd vput() calls in two places in the experimental NFS client
rmacklem [Mon, 18 Apr 2011 00:41:23 +0000 (00:41 +0000)]
Add vput() calls in two places in the experimental NFS client
that would be needed if, in the future, nfscl_loadattrcache()
were to return an error. Currently nfscl_loadattrcache()
never returns an error, so these cases never currently happen.

MFC after: 2 weeks

13 years agoChange the mutex locking for several locations in the
rmacklem [Sun, 17 Apr 2011 23:56:57 +0000 (23:56 +0000)]
Change the mutex locking for several locations in the
experimental NFS client's vnode op functions to make
them compatible with the regular NFS client. I'll admit
I'm not sure that the mutex locks around the assignments
are needed, but the regular client has them, so I added them.
Also, add handling of the case of partial attributes in
setattr to be compatible with the regular client.

MFC after: 2 weeks

13 years agoAdd checks for MNTK_UNMOUNTF at the beginning of three
rmacklem [Sun, 17 Apr 2011 23:04:03 +0000 (23:04 +0000)]
Add checks for MNTK_UNMOUNTF at the beginning of three
functions, so that threads don't get stuck in them during
a forced dismount. nfs_sync/VFS_SYNC() needs this, since it is
called by dounmount() before VFS_UNMOUNT(). The nfscl_nget()
case makes sure that a thread doing an VOP_OPEN() or
VOP_ADVLOCK() call doesn't get blocked before attempting
the RPC. Attempting RPCs don't block, since they all
fail once a forced dismount is in progress.
The third one at the beginning of nfsrpc_close()
is done so threads don't get blocked while doing VOP_INACTIVE()
as the vnodes are cleared out.
With these three changes plus a change to the umount(1)
command so that it doesn't do "sync()" for the forced case
seem to make forced dismounts work for the experimental NFS
client.

MFC after: 2 weeks

13 years agoThe change in r206686 to allow the stop argument to work for a service
dougb [Sun, 17 Apr 2011 22:31:36 +0000 (22:31 +0000)]
The change in r206686 to allow the stop argument to work for a service
that is running even though not _enable'd had an annoying side effect.
If the service was already started at boot time by another means when
the related script came around again in rcorder it would start again,
regardless of _enable, because there was a valid pid. [1]

So, split the test into 2 parts, one for (!rcvar && !stop), and one
for (stop && !valid_pid). This preserves the behavior from r206686
while preventing the undesired side effect.

PR: conf/156427 [1]
Submitted by: Eugene Grosbein <eugen@grosbein.pp.ru> [1]

13 years agokdump: Show code for signals where the default action was taken.
jilles [Sun, 17 Apr 2011 21:11:55 +0000 (21:11 +0000)]
kdump: Show code for signals where the default action was taken.

This information is available as of kernel r220740. Trace files from older
kernels will always have 0.

MFC after: 1 week

13 years agoRemove libobjc and other Objective-C related components, as these are
dim [Sun, 17 Apr 2011 21:03:23 +0000 (21:03 +0000)]
Remove libobjc and other Objective-C related components, as these are
extremely outdated, and not used by anything in the base system.

Silence from: current@

13 years agoFor clang, make -mno-mmx imply -mno-3dnow. This is what gcc does.
dim [Sun, 17 Apr 2011 20:44:02 +0000 (20:44 +0000)]
For clang, make -mno-mmx imply -mno-3dnow.  This is what gcc does.

Submitted by: arundel
Obtained from: http://llvm.org/viewvc/llvm-project?view=rev&revision=129665

13 years agoUpdate to reflect net install changes.
nwhitehorn [Sun, 17 Apr 2011 20:42:40 +0000 (20:42 +0000)]
Update to reflect net install changes.

13 years agoGet rid of the "nfscl: consider increasing kern.ipc.maxsockbuf"
rmacklem [Sun, 17 Apr 2011 20:01:32 +0000 (20:01 +0000)]
Get rid of the "nfscl: consider increasing kern.ipc.maxsockbuf"
message that was generated when doing experimental NFS client
mounts. I put that message in because the krpc would hang with
the default size for mounts that used large rsize/wsize values.
Since the bug that caused these hangs was fixed by r213756,
I think the message is no longer needed.

MFC after: 2 weeks

13 years agoFix up some of the sysctls for the experimental NFS client so
rmacklem [Sun, 17 Apr 2011 18:56:17 +0000 (18:56 +0000)]
Fix up some of the sysctls for the experimental NFS client so
that they use the same names as the regular client. Also add
string descriptions for them.

MFC after: 2 weeks

13 years agoMake dialog's --prgbox option actually work.
nwhitehorn [Sun, 17 Apr 2011 17:30:29 +0000 (17:30 +0000)]
Make dialog's --prgbox option actually work.

13 years agoUpdate dialog to version 1.1-20110302.
nwhitehorn [Sun, 17 Apr 2011 17:28:17 +0000 (17:28 +0000)]
Update dialog to version 1.1-20110302.

13 years agoUpdate dialog to 20110302 version.
nwhitehorn [Sun, 17 Apr 2011 17:00:55 +0000 (17:00 +0000)]
Update dialog to 20110302 version.

13 years agoMake in_proto.c dependent on either inet or inet6.
bz [Sun, 17 Apr 2011 16:35:16 +0000 (16:35 +0000)]
Make in_proto.c dependent on either inet or inet6.

While it does not provide any functionality for IPv6, it provides
the sysctl nodes for net.inet.* that a lot of functionality shared
between IPv4 and IPv6 depends on.  We cannot change these anymore
without breaking a lot of management and tuning.

In case of IPv6 only, we compile out everything but the sysctl node
declarations.

Reviewed by: gnn
Sponsored by: The FreeBSD Foundation
Sponsored by: iXsystems
MFC After: 5 days

13 years agoFix error where error variable was assigned result of comparison,
glebius [Sun, 17 Apr 2011 16:31:21 +0000 (16:31 +0000)]
Fix error where error variable was assigned result of comparison,
instead of function return value.

Submitted by: Przemyslaw Frasunek <przemyslaw frasunek.com>
MFC after: 4 days

13 years agoRemove hast_proto_recv(). It was used only in one place, where
trociny [Sun, 17 Apr 2011 16:18:45 +0000 (16:18 +0000)]
Remove hast_proto_recv(). It was used only in one place, where
hast_proto_recv_hdr() may be used. This also fixes the issue
(introduced by r220523) with hastctl, which crashed on assert in
hast_proto_recv_data().

Suggested and approved by: pjd (mentor)

13 years agoFix IPv6 ND. After r219562 we in nd6_ns_input() were erroneously always
bz [Sun, 17 Apr 2011 16:07:08 +0000 (16:07 +0000)]
Fix IPv6 ND.  After r219562 we in nd6_ns_input() were erroneously always
passing the cached proxydl reference (sockaddr_dl initialized or not) to
nd6_na_output().  nd6_na_output() will thus assume a proxy NA.  Revert to
conditionally passing either &proxydl or NULL if no proxy case desired.

Tested by: ipv6gw and ref9-i386
Reported by: Pete French (petefrench ingresso.co.uk on stable)
Reported by: bz, simon on Y! cluster
Reported by: kib
PR: kern/151908
MFC after: 3 days

13 years agoAllow using CMSG_NXTHDR with -Wcast-align.
jilles [Sun, 17 Apr 2011 16:04:39 +0000 (16:04 +0000)]
Allow using CMSG_NXTHDR with -Wcast-align.

If various checks are omitted, the CMSG_NXTHDR macro expands to
  (struct cmsghdr *)((char *)(cmsg) + \
  _ALIGN(((struct cmsghdr *)(cmsg))->cmsg_len))

Although there is no alignment problem (assuming cmsg is properly aligned
and _ALIGN is correct), this violates -Wcast-align on strict-alignment
architectures. Therefore an intermediate cast to void * is appropriate here.

There is no workaround other than not using -Wcast-align.

MFC after: 2 weeks

13 years agoCleanup some left-over empty directories in contrib/llvm.
dim [Sun, 17 Apr 2011 15:09:04 +0000 (15:09 +0000)]
Cleanup some left-over empty directories in contrib/llvm.

13 years agoktrace: Log the code for all signals (PSIG events).
jilles [Sun, 17 Apr 2011 14:38:11 +0000 (14:38 +0000)]
ktrace: Log the code for all signals (PSIG events).

The code provides information on how the signal was generated.

Formerly, the code was only logged for traps, much like only signal handlers
for traps received a meaningful si_code before FreeBSD 7.0.

In rare cases, no information is available and 0 is still logged.

MFC after: 1 week

13 years agoChange some defaults in the experimental NFS client to be the
rmacklem [Sun, 17 Apr 2011 14:10:12 +0000 (14:10 +0000)]
Change some defaults in the experimental NFS client to be the
same as the regular NFS client for NFSv3. The main one is making
use of a reserved port# the default. Also, set the retry limit
for TCP the same and fix the code so that it doesn't disable
readdirplus for NFSv4.

MFC after: 2 weeks

13 years agoMark the PHY as inactive before the chip is reset.
adrian [Sun, 17 Apr 2011 13:46:13 +0000 (13:46 +0000)]
Mark the PHY as inactive before the chip is reset.

It's also marked inactive by the initvals, and enabled after
the baseband/PLL has been configured, but before the RF
registers have been programmed.

The origin and reason for this particular change is currently unknown.

Obtained from: Linux ath9k

13 years agoIf we can retrieve interface address sleep for one second and try again.
sobomax [Sun, 17 Apr 2011 06:05:37 +0000 (06:05 +0000)]
If we can retrieve interface address sleep for one second and try again.
This can happen during start-up, when natd starts before dhclient has a
chance to receive IP address from the upstream provider.

MFC after: 2 weeks

13 years agoFix readdirplus in the experimental NFS client so that it
rmacklem [Sun, 17 Apr 2011 02:44:51 +0000 (02:44 +0000)]
Fix readdirplus in the experimental NFS client so that it
skips over ".." to avoid a LOR race with nfs_lookup(). This
fix is analagous to r138256 in the regular NFS client.

MFC after: 2 weeks

13 years agoAdd a lktype flags argument to nfscl_nget() and ncl_nget() in the
rmacklem [Sat, 16 Apr 2011 23:20:21 +0000 (23:20 +0000)]
Add a lktype flags argument to nfscl_nget() and ncl_nget() in the
experimental NFS client so that its nfs_lookup() function can use
cn_lkflags in a manner analagous to the regular NFS client.

MFC after: 2 weeks

13 years agoAdd mutex locking on the nfs node in ncl_inactive() for the
rmacklem [Sat, 16 Apr 2011 22:15:59 +0000 (22:15 +0000)]
Add mutex locking on the nfs node in ncl_inactive() for the
experimental NFS client.

MFC after: 2 weeks

13 years agoRemove malloc(9) return value checks when M_WAITOK is used.
dchagin [Sat, 16 Apr 2011 16:20:51 +0000 (16:20 +0000)]
Remove malloc(9) return value checks when M_WAITOK is used.

MFC after: 2 Week

13 years agoAdd some new features:
bschmidt [Sat, 16 Apr 2011 14:56:13 +0000 (14:56 +0000)]
Add some new features:
- 6000 series devices need enhanced sensitivity calibration.
- 6000 series devices need a different setting for the shadow reg.
- set the IWN_FLAG_HAS_11N bit if the EEPROM says the device has 11n
  support.

Obtained from: OpenBSD

13 years agoBring over the HAL/OPS changes, instead of two const structs it is now
bschmidt [Sat, 16 Apr 2011 14:51:46 +0000 (14:51 +0000)]
Bring over the HAL/OPS changes, instead of two const structs it is now
slightly more dynamic.

Obtained from: OpenBSD

13 years agoRead RX/TX chainmasks directly of the EEPROM. Some chips are known to
bschmidt [Sat, 16 Apr 2011 14:49:29 +0000 (14:49 +0000)]
Read RX/TX chainmasks directly of the EEPROM. Some chips are known to
have the wrong/broken information stored, keep the hardcoded values for
those.

Obtained from: OpenBSD

13 years agoJust a whitespace sync, some more style(9) conform then others.
bschmidt [Sat, 16 Apr 2011 14:40:42 +0000 (14:40 +0000)]
Just a whitespace sync, some more style(9) conform then others.

13 years agoSync comments with OpenBSD.
bschmidt [Sat, 16 Apr 2011 14:38:57 +0000 (14:38 +0000)]
Sync comments with OpenBSD.

13 years agoSync debug and error messages with OpenBSD.
bschmidt [Sat, 16 Apr 2011 14:38:01 +0000 (14:38 +0000)]
Sync debug and error messages with OpenBSD.

13 years agoShuffle code around a bit. Mostly to group functional connected things,
bschmidt [Sat, 16 Apr 2011 14:36:43 +0000 (14:36 +0000)]
Shuffle code around a bit. Mostly to group functional connected things,
others to get the same order as the OpenBSD code.

13 years agoDon't do Kite antenna switch selection this way (for now); antenna
adrian [Sat, 16 Apr 2011 13:47:17 +0000 (13:47 +0000)]
Don't do Kite antenna switch selection this way (for now); antenna
diversity is done elsewhere now.

13 years agoRename some stuff in favour of the OpenBSD names:
bschmidt [Sat, 16 Apr 2011 13:40:32 +0000 (13:40 +0000)]
Rename some stuff in favour of the OpenBSD names:
- prefer EDCA over WME
- qid for a TXQ ID
- reg for register values

13 years agoFix WME/QoS handling:
bschmidt [Sat, 16 Apr 2011 13:25:19 +0000 (13:25 +0000)]
Fix WME/QoS handling:
- move the TX queue selection into iwn_tx_data/iwn_tx_data_raw
- extract traffic identifier and use it
- do not expect ACKs for frames marked as such

13 years agoRemove if_ierrors which do not necessarily indicate a RX error, also
bschmidt [Sat, 16 Apr 2011 13:12:12 +0000 (13:12 +0000)]
Remove if_ierrors which do not necessarily indicate a RX error, also
do account send packets. While here use the IWN_TX_FAIL constant.

13 years agoDisable classic-style fast diversity on the AR5416 and later.
adrian [Sat, 16 Apr 2011 12:46:46 +0000 (12:46 +0000)]
Disable classic-style fast diversity on the AR5416 and later.

Antenna diversity on the >= AR5416 is implemented differently than the
AR5212 and previous chips. So for now, and not to confuse things, just
disable it for now.

13 years agoInstead of hardcoding TX rates and using that to fill the retry table
bschmidt [Sat, 16 Apr 2011 12:42:54 +0000 (12:42 +0000)]
Instead of hardcoding TX rates and using that to fill the retry table
use the neogotiated ni_rates instead.

13 years agoRemove some duplicate code from the AR9285 TX power configuration path.
adrian [Sat, 16 Apr 2011 11:59:37 +0000 (11:59 +0000)]
Remove some duplicate code from the AR9285 TX power configuration path.

13 years agoRemove missing include directory in preparation for adding
brucec [Sat, 16 Apr 2011 11:15:57 +0000 (11:15 +0000)]
Remove missing include directory in preparation for adding
-Wmissing-include-dirs to CWARNFLAGS.

13 years agoReplace RX/TX ring allocation error messages with something more sane
bschmidt [Sat, 16 Apr 2011 11:14:42 +0000 (11:14 +0000)]
Replace RX/TX ring allocation error messages with something more sane
and remove those where the caller already prints one.

13 years agoPrevent double-free, also use the same error codes as OpenBSD.
bschmidt [Sat, 16 Apr 2011 11:11:22 +0000 (11:11 +0000)]
Prevent double-free, also use the same error codes as OpenBSD.

13 years agoAdd missing bus_dmamap_sync calls as well as remove two duplicate ones.
bschmidt [Sat, 16 Apr 2011 10:52:11 +0000 (10:52 +0000)]
Add missing bus_dmamap_sync calls as well as remove two duplicate ones.

Obtained from: OpenBSD

13 years agoUnify TX/RX ring allocation, finish the descriptior DMA stuff before
bschmidt [Sat, 16 Apr 2011 10:38:27 +0000 (10:38 +0000)]
Unify TX/RX ring allocation, finish the descriptior DMA stuff before
starting with data.

13 years agoMake sure to destroy all DMA tags and maps.
bschmidt [Sat, 16 Apr 2011 10:35:02 +0000 (10:35 +0000)]
Make sure to destroy all DMA tags and maps.

13 years agoRewrite DMA segment handling to be more inline with the OpenBSD code.
bschmidt [Sat, 16 Apr 2011 10:32:46 +0000 (10:32 +0000)]
Rewrite DMA segment handling to be more inline with the OpenBSD code.
Also change the m_len == 0 hack to have less code churn.

13 years agoscratch_paddr has the same address pre-assigned, use that instead.
bschmidt [Sat, 16 Apr 2011 09:39:08 +0000 (09:39 +0000)]
scratch_paddr has the same address pre-assigned, use that instead.

13 years agoIn case a new mbuf can't be loaded, reuse the old one.
bschmidt [Sat, 16 Apr 2011 09:34:40 +0000 (09:34 +0000)]
In case a new mbuf can't be loaded, reuse the old one.

13 years agoOpenBSD uses IWN_RBUF_SIZE not MJUMPAGESIZE for the RX path, also replace
bschmidt [Sat, 16 Apr 2011 08:59:21 +0000 (08:59 +0000)]
OpenBSD uses IWN_RBUF_SIZE not MJUMPAGESIZE for the RX path, also replace
caddr_t with void * to be in sync.

13 years agoRemove the flags argument of iwn_dma_contig_alloc(), it is always set
bschmidt [Sat, 16 Apr 2011 08:57:17 +0000 (08:57 +0000)]
Remove the flags argument of iwn_dma_contig_alloc(), it is always set
as BUS_DMA_NOWAIT. While here also set BUS_DMA_COHERENT.

13 years agoMove a comment to the right place. The rearrange done in r208928 left
jh [Sat, 16 Apr 2011 08:38:11 +0000 (08:38 +0000)]
Move a comment to the right place. The rearrange done in r208928 left
the comment to a wrong place.

13 years agoRSSI related syncs with the OpenBSD code:
bschmidt [Sat, 16 Apr 2011 07:32:57 +0000 (07:32 +0000)]
RSSI related syncs with the OpenBSD code:
- read RSSI only for the active chains
- cast RSSI/NF to int8_t before passing it up to radiotap
- remove the htole64() for the timestamp

Obtained from: OpenBSD

13 years agoPass errors that might happen during state transitions up to net80211.
bschmidt [Sat, 16 Apr 2011 07:21:00 +0000 (07:21 +0000)]
Pass errors that might happen during state transitions up to net80211.

13 years agoObtain the channel number directly from the laster RXON command, also
bschmidt [Sat, 16 Apr 2011 07:17:03 +0000 (07:17 +0000)]
Obtain the channel number directly from the laster RXON command, also
chan is an uint8_t.

13 years agoMove TUNABLE_INT_FETCH() out of the SIM lock to avoid sleeping while
jh [Sat, 16 Apr 2011 06:54:41 +0000 (06:54 +0000)]
Move TUNABLE_INT_FETCH() out of the SIM lock to avoid sleeping while
holding the lock. The fix is analogous to r220618 for ada(4).

Reviewed by: mav

13 years agoMFi386: revisions 220389 and 220392
nyan [Sat, 16 Apr 2011 06:20:49 +0000 (06:20 +0000)]
MFi386: revisions 220389 and 220392

  - Mark getc() as inline, this has no effect on gcc but helps clang.
  - Move getc() body before xgetc() so gcc does not emit a warning about
    function having no body.

13 years agoChange the experimental NFS client so that it creates nfsiod
rmacklem [Fri, 15 Apr 2011 23:07:48 +0000 (23:07 +0000)]
Change the experimental NFS client so that it creates nfsiod
threads in the same manner as the regular NFS client after
r214026 was committed. This resolves the lors fixed by r214026
and its predecessors for the regular client.

Reviewed by: jhb
MFC after: 2 weeks

13 years agoPurge _USE_BERKELEY_YACC definition from Makefile. This ugly hack is no
jkim [Fri, 15 Apr 2011 21:47:10 +0000 (21:47 +0000)]
Purge _USE_BERKELEY_YACC definition from Makefile.  This ugly hack is no
long necessary for us since r220680.

13 years agoRe-merge with ACPICA vendor source.
jkim [Fri, 15 Apr 2011 21:38:24 +0000 (21:38 +0000)]
Re-merge with ACPICA vendor source.

13 years agofix the order of the prototypes from the previosu commit
bschmidt [Fri, 15 Apr 2011 20:40:49 +0000 (20:40 +0000)]
fix the order of the prototypes from the previosu commit