]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years agoMove opt_apic.h closer to the "XXX: for assym.s" comment.
des [Tue, 26 May 2009 20:57:19 +0000 (20:57 +0000)]
Move opt_apic.h closer to the "XXX: for assym.s" comment.

Suggested by: jhb
MFC after: 1 week

15 years agoAdd the OpenSolaris dtrace lockstat provider. The lockstat provider
sson [Tue, 26 May 2009 20:28:22 +0000 (20:28 +0000)]
Add the OpenSolaris dtrace lockstat provider.  The lockstat provider
adds probes for mutexes, reader/writer and shared/exclusive locks to
gather contention statistics and other locking information for
dtrace scripts, the lockstat(1M) command and other potential
consumers.

Reviewed by: attilio jhb jb
Approved by: gnn (mentor)

15 years agoopt_apic.h is i386-only.
des [Tue, 26 May 2009 20:22:03 +0000 (20:22 +0000)]
opt_apic.h is i386-only.

MFC after: 1 week

15 years agoCopy ee 1.4.2 into the contrib directory.
ed [Tue, 26 May 2009 20:13:17 +0000 (20:13 +0000)]
Copy ee 1.4.2 into the contrib directory.

This allows me to merge our custom changes to ee(1) back on top of
original sources, with correct mergeinfo.

15 years agoAdd a regression test for kern/21768.
des [Tue, 26 May 2009 20:13:06 +0000 (20:13 +0000)]
Add a regression test for kern/21768.

MFC after: 1 week

15 years agoCorrect the sense of a test so that this filter always waits for the full
jhb [Tue, 26 May 2009 20:00:30 +0000 (20:00 +0000)]
Correct the sense of a test so that this filter always waits for the full
request to arrive.  Previously it would end up returning as soon as the
request length stored in the first two bytes had arrived.

Reviewed by: dwmalone
MFC after: 1 week

15 years agoRemove unused curses replacement implementation from ee(1) sources.
ed [Tue, 26 May 2009 19:06:56 +0000 (19:06 +0000)]
Remove unused curses replacement implementation from ee(1) sources.

15 years agoVendor import of ee 1.4.2.
ed [Tue, 26 May 2009 18:50:01 +0000 (18:50 +0000)]
Vendor import of ee 1.4.2.

15 years agoGet rid of M_TEMP.
ed [Tue, 26 May 2009 18:33:36 +0000 (18:33 +0000)]
Get rid of M_TEMP.

15 years agoDo not forcefully close the write transfer when closing the tty, it needs to
thompsa [Tue, 26 May 2009 17:06:36 +0000 (17:06 +0000)]
Do not forcefully close the write transfer when closing the tty, it needs to
run to completion and drain the tty queue.

15 years agoAdapt to the new ACL #define names.
trasz [Tue, 26 May 2009 17:01:00 +0000 (17:01 +0000)]
Adapt to the new ACL #define names.

Reviewed by: rmacklem@

15 years agoAdd two new utilities and two new daemons to /usr/src/usr.sbin that
rmacklem [Tue, 26 May 2009 15:19:04 +0000 (15:19 +0000)]
Add two new utilities and two new daemons to /usr/src/usr.sbin that
are specifically used by the experimental nfsv4 subsystem.
  nfscbd - The NFSv4 client callback daemon.
  nfsuserd - The NFSv4 daemon that maps between user and group name
           and their corresponding uid/gid numbers.
  nfsdumpstate - A utility that dumps out the NFSv4 Open/Lock state.
  nfsrevoke - Administratively revokes an NFSv4 client, releasing all
           NFSv4 Open/Lock state it holds on the server.

Approved by: kib (mentor)

15 years ago- Unbreak 64 bit platforms by casting off_t to intmax.
lulf [Tue, 26 May 2009 14:15:06 +0000 (14:15 +0000)]
- Unbreak 64 bit platforms by casting off_t to intmax.

15 years agoUnbreak bktr(4).
ed [Tue, 26 May 2009 12:01:37 +0000 (12:01 +0000)]
Unbreak bktr(4).

15 years agoChange license to more bori^Wadul^Wcanonical.
trasz [Tue, 26 May 2009 11:42:06 +0000 (11:42 +0000)]
Change license to more bori^Wadul^Wcanonical.

Submitted by: rwatson@

15 years ago- Fix wrong print on BIO_DONE.
lulf [Tue, 26 May 2009 10:03:44 +0000 (10:03 +0000)]
- Fix wrong print on BIO_DONE.
- Use db_printf instead of printf. While here, apply this to other ddb commands
  as well.

Pointed out by: pjd

15 years agoAdd missing socket options.
pjd [Tue, 26 May 2009 09:19:21 +0000 (09:19 +0000)]
Add missing socket options.

15 years agoMFp4 changes neccessary for NFSv4 ACLs support in ZFS. This is mostly
trasz [Tue, 26 May 2009 08:21:59 +0000 (08:21 +0000)]
MFp4 changes neccessary for NFSv4 ACLs support in ZFS.  This is mostly
about removing a few #ifdefs and providing compatibility wrappers and
VOP implementations to get and set an ACL; ZFS does ACL enforcement all
by itself.

Note that the VOPs are ifdefed out for now, so this change should be
a no-op.

Reviewed by: pjd

15 years agoMention the danger of running programs using ``!''.
brian [Tue, 26 May 2009 07:40:32 +0000 (07:40 +0000)]
Mention the danger of running programs using ``!''.

PR: 112481
MFC after: 1 week

15 years agoFix a race that can stall the timer when we remove a timer that has another
brian [Tue, 26 May 2009 07:32:08 +0000 (07:32 +0000)]
Fix a race that can stall the timer when we remove a timer that has another
timer with a <0.05 second delta next to it.

This is done by avoiding the possibility of updating the first residual
time delta in the timer list to zero.

PR: 102747
Submitted by: Sergey Zaharchenko - doublef-ctm at yandex dot ru
MFC after: 3 weeks

15 years ago- Add 'show bio' DDB command.
lulf [Tue, 26 May 2009 07:29:17 +0000 (07:29 +0000)]
- Add 'show bio' DDB command.

MFC after: 3 weeks

15 years agoSet PG_WRITEABLE in Book-E pmap_enter[_locked] if it creates a mapping that
raj [Tue, 26 May 2009 06:24:50 +0000 (06:24 +0000)]
Set PG_WRITEABLE in Book-E pmap_enter[_locked] if it creates a mapping that
permits write access. This is similar to r192671.

Pointed out and reviewed by: alc

15 years agoFree the memory correctly in the error case
benjsc [Tue, 26 May 2009 03:56:56 +0000 (03:56 +0000)]
Free the memory correctly in the error case

Submitted by: frtzkatz at yahoo.com
Approved by: sam
MFC after: 1 month

15 years agoAdd two sysctl variables to the experimental nfs server, so
rmacklem [Tue, 26 May 2009 01:47:37 +0000 (01:47 +0000)]
Add two sysctl variables to the experimental nfs server, so
that the range of versions of NFS handled by the server can
be limited. The nfsd daemon must be restarted after these
sysctl variables are changed, in order for the change to take
effect.

Approved by: kib (mentor)

15 years agoFix the handling of NFSv4 Illegal Operation number to conform
rmacklem [Tue, 26 May 2009 01:16:09 +0000 (01:16 +0000)]
Fix the handling of NFSv4 Illegal Operation number to conform
to RFC3530 (the operation number in the reply must be set to
the value for OP_ILLEGAL). Also cleaned up some indentation.

Approved by: kib (mentor)

15 years agoFix the experimental nfs server's interface to the new krpc so
rmacklem [Tue, 26 May 2009 01:09:33 +0000 (01:09 +0000)]
Fix the experimental nfs server's interface to the new krpc so
that it handles the case of a non-exported NFSv4 root correctly.
Also, delete handling for the case where nd_repstat is already
set in nfs_proc(), since that no longer happens.

Approved by: kib (mentor)

15 years agoPer the man page:
dougb [Tue, 26 May 2009 00:07:02 +0000 (00:07 +0000)]
Per the man page:
The makekey utility has been deprecated and will be removed in a future
release of FreeBSD.

Actually removing it was approved back on 10/29/2007 by re (kensmith) but
I dropped the ball on actually removing it. It's doubtful that it's become
more relevant/useful in the intervening time.

15 years agoI have been here long
markm [Mon, 25 May 2009 23:06:46 +0000 (23:06 +0000)]
I have been here long
The duration I know not
I tell what I can

15 years agoThere is rubbish here
markm [Mon, 25 May 2009 22:50:11 +0000 (22:50 +0000)]
There is rubbish here
It is time to take it out
Now it is cleaner

15 years agoBump the document date to reflect the recent address range enhancements.
brian [Mon, 25 May 2009 21:29:06 +0000 (21:29 +0000)]
Bump the document date to reflect the recent address range enhancements.

Suggested by: throdes

15 years agoBump the document date to reflect the 'p' command enhancements.
brian [Mon, 25 May 2009 21:27:31 +0000 (21:27 +0000)]
Bump the document date to reflect the 'p' command enhancements.

Suggested by: trhodes

15 years agoFix kdump build when building it by hand.
ed [Mon, 25 May 2009 20:07:41 +0000 (20:07 +0000)]
Fix kdump build when building it by hand.

I don't entirely like this approach, but it will only be temporarily,
namely until we get rid of COMPAT_43TTY. I do want <sys/ioctl_compat.h>
to cause a compiler error when included, because it's just there for
binary compatibility.

Reported by: Andrzej Tobola <ato iem pw edu pl>

15 years agoEuro is missing
markm [Mon, 25 May 2009 17:06:24 +0000 (17:06 +0000)]
Euro is missing
With this I shall see it fixed
Thank you for your time

15 years agoFix handling of devices w/o radiotap support:
sam [Mon, 25 May 2009 16:38:47 +0000 (16:38 +0000)]
Fix handling of devices w/o radiotap support:
o do not attach DLT_IEEE802_11_RADIO unless both tx and rx headers are
  present; this is assumed in the capture code paths
o verify the above with asserts in ieee80211_radiotap_{rx,tx}
o add missing checks for active taps before calling ieee80211_radiotap_rx

15 years agodo internal bpf tracking only for radiotap DLT's
sam [Mon, 25 May 2009 16:35:31 +0000 (16:35 +0000)]
do internal bpf tracking only for radiotap DLT's

15 years agorev bpf attach/detach event api to include the dlt
sam [Mon, 25 May 2009 16:34:35 +0000 (16:34 +0000)]
rev bpf attach/detach event api to include the dlt

15 years agoAdd the "-4" option to nfsstat, so that it will reports stats
rmacklem [Mon, 25 May 2009 15:48:10 +0000 (15:48 +0000)]
Add the "-4" option to nfsstat, so that it will reports stats
for the experimental nfs subsystem when that option is specified.

Approved by: kib (mentor)

15 years agoRemove comment about moving tcp_reass() to its own file named tcp_reass.c,
rwatson [Mon, 25 May 2009 14:51:47 +0000 (14:51 +0000)]
Remove comment about moving tcp_reass() to its own file named tcp_reass.c,
that happened a while ago.

MFC after: 3 days

15 years agoUse, in uncovered part, the END() macro in order to improve debugging.
attilio [Mon, 25 May 2009 14:37:10 +0000 (14:37 +0000)]
Use, in uncovered part, the END() macro in order to improve debugging.
In this specific case, Valgrind won't get confused when analyzing such
functions.

Sponsored by: Sandvine Incorporated
Tested by: emaste
MFC: 3 days

15 years agoStaticize spx_remque() now that it's only used from spx_reass.c.
rwatson [Mon, 25 May 2009 13:52:51 +0000 (13:52 +0000)]
Staticize spx_remque() now that it's only used from spx_reass.c.

15 years agoAdd missing call to ipx_pcbdetach() during SPX socket tear-down: not
rwatson [Mon, 25 May 2009 13:34:03 +0000 (13:34 +0000)]
Add missing call to ipx_pcbdetach() during SPX socket tear-down: not
harmful in practice if running without INVARIANTS, but will panic
with KASSERT enabled when SPX sockets are closed.

MFC after: 3 days

15 years agoEliminate use of dtom() in spx_output() by fixing up tracking of the
rwatson [Mon, 25 May 2009 13:32:54 +0000 (13:32 +0000)]
Eliminate use of dtom() in spx_output() by fixing up tracking of the
containing mbuf for 'si' in local variable 'm'.

MFC after: 1 month

15 years agoPrefer NULL to 0 for pointer assignments.
rwatson [Mon, 25 May 2009 11:52:33 +0000 (11:52 +0000)]
Prefer NULL to 0 for pointer assignments.

MFC after: 1 month

15 years agoRather than store a skeleton IPX header in an mbuf hung off the SPX
rwatson [Mon, 25 May 2009 11:50:58 +0000 (11:50 +0000)]
Rather than store a skeleton IPX header in an mbuf hung off the SPX
PCB, simply embed it in the PCB, avoiding additional memory overhead,
memory allocation overhead, and removing one of the few remaining
uses of dtom() in the network stack.

Restore misplaced spx_ctlinput() from an earlier commit.

MFC after: 1 month

15 years agoPull SPX reassembly queue init and flush into spx_reass.c.
rwatson [Mon, 25 May 2009 10:25:41 +0000 (10:25 +0000)]
Pull SPX reassembly queue init and flush into spx_reass.c.

MFC after: 1 month

15 years agoPrefer m_nextpkt to m_act when iterating mbuf queues.
rwatson [Mon, 25 May 2009 09:57:18 +0000 (09:57 +0000)]
Prefer m_nextpkt to m_act when iterating mbuf queues.

MFC after: 1 month

15 years agoo Fix typo in the example.
maxim [Mon, 25 May 2009 09:52:58 +0000 (09:52 +0000)]
o Fix typo in the example.

PR: docs/134930
Submitted by: Alex Keda
MFC after: 1 week

15 years agoComplete move of SPX reassembly from spx_usrreq.c to spx_reass.c.
rwatson [Mon, 25 May 2009 09:28:54 +0000 (09:28 +0000)]
Complete move of SPX reassembly from spx_usrreq.c to spx_reass.c.

MFC after: 1 month

15 years agoEnhance the 'p' command so that it understands size qualifiers (K/M/G) and
brian [Mon, 25 May 2009 09:23:26 +0000 (09:23 +0000)]
Enhance the 'p' command so that it understands size qualifiers (K/M/G) and
so that it understands '*' as 'DTRT'.

PR: 68312
Submitted by: Rene de Vries - rene at tunix dot nl (mostly)
MFC after: 3 weeks

15 years agoCopy spx_usrreq.c to spx_reass.c in order to apply similar file layout
rwatson [Mon, 25 May 2009 09:09:43 +0000 (09:09 +0000)]
Copy spx_usrreq.c to spx_reass.c in order to apply similar file layout
changes to IPX/SPX that were applied to TCP/IP in the creation of
tcp_reass.c.

MFC after: 1 month

15 years agoCorrect controller description for 88E8035, 88E8036, 88E8038 and
yongari [Mon, 25 May 2009 08:27:52 +0000 (08:27 +0000)]
Correct controller description for 88E803588E803688E8038 and
88E8039. These are fast ethernet controllers.

15 years agoCorrect controller description for 88E8035, 88E8036, 88E8038 and
yongari [Mon, 25 May 2009 08:26:39 +0000 (08:26 +0000)]
Correct controller description for 88E803588E803688E8038 and
88E8039. These are fast ethernet controllers.

15 years agomsk(4) now supports Yukon FE+. Specifically 88E8040, 88E8040T,
yongari [Mon, 25 May 2009 08:11:03 +0000 (08:11 +0000)]
msk(4) now supports Yukon FE+. Specifically 88E8040, 88E8040T,
88E8048 and 88E8070 are supported.
Bump .Dd

15 years agoSimplify SIOCSIFFLAGS handler.
yongari [Mon, 25 May 2009 08:02:05 +0000 (08:02 +0000)]
Simplify SIOCSIFFLAGS handler.

15 years agoBe consistent with other capability checking.
yongari [Mon, 25 May 2009 07:59:30 +0000 (07:59 +0000)]
Be consistent with other capability checking.

15 years agoDon't reinitialize controller when interface is already running.
yongari [Mon, 25 May 2009 07:56:14 +0000 (07:56 +0000)]
Don't reinitialize controller when interface is already running.

15 years agoIf interface is not UP, don't return media status.
yongari [Mon, 25 May 2009 07:50:14 +0000 (07:50 +0000)]
If interface is not UP, don't return media status.

15 years agoAdd device ids for Yukon FE+(88E8040, 88E8040T, 88E8048 and 88E8070).
yongari [Mon, 25 May 2009 07:48:00 +0000 (07:48 +0000)]
Add device ids for Yukon FE+(88E8040, 88E8040T, 88E8048 and 88E8070).

15 years agoAdd workaround for Yukon FE+ A0. This controller is known to have
yongari [Mon, 25 May 2009 07:31:18 +0000 (07:31 +0000)]
Add workaround for Yukon FE+ A0. This controller is known to have
severe silicon bugs that can't handle VLAN hardware tagging as well
as status LE writeback bug. The status LE writeback bug is so
critical we can't trust status word of received frame. To accept
frames on Yukon FE+ A0 msk(4) just do minimal check for received
frames and pass them to upper stack. This means msk(4) can pass
corrupted frames to upper layer. You have been warned!
Also I supposed RX_GMF_FL_THR to be 32bits register but Linux
driver treated it as 16bit register so follow their leads. At least
this does not seem to break msk(4) on Yukon FE+.

Tested by: bz, Tanguy Bouzeloc ( the.zauron <> gmail dot com )
Bruce Cran ( bruce <> cran dot org dot uk )
Michael Reifenberger ( mike <> reifenberger dot com )
Stephen Montgomery-Smith ( stephen <> missouri dot edu )

15 years agoAdd preliminary Yukon FE+ support and register definitions.
yongari [Mon, 25 May 2009 07:06:10 +0000 (07:06 +0000)]
Add preliminary Yukon FE+ support and register definitions.
Yukon FE+ is fast ethernet controller and uses new descriptor
format. Since I don't have this controller, the support code was
written from guess and various feedback from enthusiastic users.
Thanks to all users who patiently tested my initial patches.
Special thanks to Tanguy Bouzeloc who fixed critical bug of initial
patch.

Tested by: bz, Tanguy Bouzeloc ( the.zauron <> gmail dot com )
Bruce Cran ( bruce <> cran dot org dot uk )
Michael Reifenberger ( mike <> reifenberger dot com )
Stephen Montgomery-Smith ( stephen <> missouri dot edu )

15 years agoRegression test the 'addr1,+N' feature added in r192732
brian [Mon, 25 May 2009 06:58:42 +0000 (06:58 +0000)]
Regression test the 'addr1,+N' feature added in r192732

15 years agoImplement "addr1,+N" ranges - not dissimilar to grep's -A switch.
brian [Mon, 25 May 2009 06:45:33 +0000 (06:45 +0000)]
Implement "addr1,+N" ranges - not dissimilar to grep's -A switch.

PR: 134856
Submitted by: Jeremie Le Hen - jeremie at le-hen dot org

15 years agoExplicitly reset GMAC Controls and initialize GM_GP_CTRL register.
yongari [Mon, 25 May 2009 06:39:48 +0000 (06:39 +0000)]
Explicitly reset GMAC Controls and initialize GM_GP_CTRL register.
The GM_GP_CTRL register may have stale content from previous link
information so clearing it will make hardware update the register
correctly when it established a valid link.
While I'm here remove stale comment.

15 years agoDisable HW WOL for Yukon EC Ultra. While I'm here use switch
yongari [Mon, 25 May 2009 06:29:02 +0000 (06:29 +0000)]
Disable HW WOL for Yukon EC Ultra. While I'm here use switch
statement over if-else statement. This change will make it easy to
add newer Yukon controllers.

15 years agoExplicitly check resolved speed/duplex. Just checking IFM_ACTIVE
yongari [Mon, 25 May 2009 06:19:36 +0000 (06:19 +0000)]
Explicitly check resolved speed/duplex. Just checking IFM_ACTIVE
does not guarantee established link. Also 1000baseT link report for
fast ethernet controller is not valid one so make sure gigabit link
is allowed for this controller.
Whenever we lost link, check whether Rx/Tx MACs were enabled. If both
MAC are not active, do not try to disable it again.

15 years agoAdd support for newer descriptor format. This format is used on
yongari [Mon, 25 May 2009 06:09:18 +0000 (06:09 +0000)]
Add support for newer descriptor format. This format is used on
Yukon FE+, Yukon Extreme and Yukon Supreme.

15 years agoOops, add missing ~ operator.
yongari [Mon, 25 May 2009 04:27:12 +0000 (04:27 +0000)]
Oops, add missing ~ operator.

15 years agoCaller already hold a driver lock in mii callback, assert it.
yongari [Mon, 25 May 2009 04:25:08 +0000 (04:25 +0000)]
Caller already hold a driver lock in mii callback, assert it.

15 years agoRather than checking every chip revision, introduce more flags to
yongari [Mon, 25 May 2009 04:22:27 +0000 (04:22 +0000)]
Rather than checking every chip revision, introduce more flags to
mark controller's capability. Controllers that have jumbo frame
support sets MSK_FLAG_JUMBO, and controllers that does not support
checksum offloading for jumbo frames will set MSK_FLAG_JUMBO_NOCSUM.
For Fast Ethernet controllers it will set MSK_FLAG_FASTETHER and it
would be used in link state handling.

While here, disable Tx checksum offloading if jumbo frame is used
on controllers that does not have Tx checksum offloading capability
for jumbo frame(e.g. Yukon EC Ultra).

15 years agoCorrectly return the result of mii_mediachg(). Previously it always
yongari [Mon, 25 May 2009 03:53:12 +0000 (03:53 +0000)]
Correctly return the result of mii_mediachg(). Previously it always
used to return success.

15 years agoUse bit definition to represent MSI and detach state instead of
yongari [Mon, 25 May 2009 03:49:43 +0000 (03:49 +0000)]
Use bit definition to represent MSI and detach state instead of
using separate variables in softc.

15 years agoUse bit definition to represent link state, device suspend instead
yongari [Mon, 25 May 2009 03:42:33 +0000 (03:42 +0000)]
Use bit definition to represent link state, device suspend instead
of using separate variables in softc.

15 years agoRemove link handling taskqueue and use mii callback directly. While
yongari [Mon, 25 May 2009 03:24:47 +0000 (03:24 +0000)]
Remove link handling taskqueue and use mii callback directly. While
I'm here also checks driver running state.

15 years agoAdd driver support for 88E3016 PHY which is found on Marvell Yukon
yongari [Mon, 25 May 2009 02:36:29 +0000 (02:36 +0000)]
Add driver support for 88E3016 PHY which is found on Marvell Yukon
FE+ controller. Due to the severe silicon bugs for Yukon FE+,
88E3016 seems to require more workarounds. However I'm not sure
whether the workaround is PHY specific or only applicable to Yukon
FE+. The datasheet for the PHY is publicly available but it lacks
several details for the workaround used in this change. The
workaround information was obtained from Linux. Many thanks to
Yukon FE+ users who helped me add 88E3016 support.

Tested by: bz, Tanguy Bouzeloc ( the.zauron <> gmail dot com )
Bruce Cran ( bruce <> cran dot org dot uk )
Michael Reifenberger ( mike <> reifenberger dot com )
Stephen Montgomery-Smith ( stephen <> missouri dot edu )

15 years agoDo not ignore NEXT Page capability of auto-negotiation
yongari [Mon, 25 May 2009 02:05:00 +0000 (02:05 +0000)]
Do not ignore NEXT Page capability of auto-negotiation
advertisement register. Some PHYs such as 88E3016 requires NEXT
Page capability to establish valid link. Also set protocol selector
field which is read only but it makes the intention clearer.

15 years agoDon't read unnecessary PHY registers. Speed/duplex resolution bit
yongari [Mon, 25 May 2009 01:56:19 +0000 (01:56 +0000)]
Don't read unnecessary PHY registers. Speed/duplex resolution bit
is valid only for auto-negotiation case so check the bit if we know
auto-negotiation is active. While I'm here explicitly checks
current speed with speed mask and set IFM_NONE if resolved speed
is unknown.

15 years agoReport current link state while auto-negotiation is in progress.
yongari [Mon, 25 May 2009 01:45:28 +0000 (01:45 +0000)]
Report current link state while auto-negotiation is in progress.

15 years agoUse mii_phy_add_media() and remove usage of local macro ADD. Also
yongari [Mon, 25 May 2009 01:41:05 +0000 (01:41 +0000)]
Use mii_phy_add_media() and remove usage of local macro ADD. Also
checks extended status register to see whether the PHY is fast
ethernet or not. This removes a lot of checks for specific PHY
models and it makes easy to add more PHYs to e1000phy(4).

While I'm here remove setting mii_anegticks as it is set with
mii_phy_add_media().

15 years agoAdd NFSv4 root export checks to the DelegPurge, Renew and
rmacklem [Mon, 25 May 2009 01:00:09 +0000 (01:00 +0000)]
Add NFSv4 root export checks to the DelegPurge, Renew and
ReleaseLockOwner operations analagous to what is already
in place for SetClientID and SetClientIDConfirm. These are
the five NFSv4 operations that do not use file handle(s),
so the checks are done using the NFSv4 root export entries
in /etc/exports.

Approved by: kib (mentor)

15 years agoNVIDIA MCP controllers have no Rx buffer alignment restrictions.
yongari [Mon, 25 May 2009 00:56:01 +0000 (00:56 +0000)]
NVIDIA MCP controllers have no Rx buffer alignment restrictions.
Remove PAGE_SIZE alignment used in Rx buffer DMA tag creation. The
alignment restriction was used in old local jumbo allocator and
nfe(4) switched to UMA backed page allocator for jumbo frame.

This change should fix jumbo buffer allocation failure.

Reported by: Pascal Braun ( pascal.braun <> continum dot net )

15 years agoTemporarily #undef NFS4_ACL_EXTATTR_NAME, so that the
rmacklem [Sun, 24 May 2009 23:47:22 +0000 (23:47 +0000)]
Temporarily #undef NFS4_ACL_EXTATTR_NAME, so that the
experimental nfs subsystem will build while the NFSv4 ACL
support is going into the kernel.

Approved by: kib (mentor)

15 years agoThere are things too complex to be fixed in one commit.
trasz [Sun, 24 May 2009 20:34:29 +0000 (20:34 +0000)]
There are things too complex to be fixed in one commit.
Fix a typo in acl(9) manual page.

Submitted by: avg

15 years agoCrib the realign function out of nfs_krpc.c and add a call
rmacklem [Sun, 24 May 2009 19:46:12 +0000 (19:46 +0000)]
Crib the realign function out of nfs_krpc.c and add a call
to it for the client side reply. Hopefully this fixes the
problem with using the new krpc for arm for the experimental
nfs client.

Approved by: kib (mentor)

15 years agoDon't allow non-owner to set SUID bit on a file. It doesn't make
trasz [Sun, 24 May 2009 19:21:49 +0000 (19:21 +0000)]
Don't allow non-owner to set SUID bit on a file.  It doesn't make
any difference now, but in NFSv4 ACLs, there is write_acl permission,
which also affects mode changes.

Reviewed by: pjd

15 years agoFix the experimental NFSv4 server so that it handles the case
rmacklem [Sun, 24 May 2009 18:49:53 +0000 (18:49 +0000)]
Fix the experimental NFSv4 server so that it handles the case
where a client is not allowed NFSv4 access correctly. This
restriction is specified in the "V4: ..." line(s) in
/etc/exports.

Approved by: kib (mentor)

15 years agoRemove an unused variable.
antoine [Sun, 24 May 2009 18:35:53 +0000 (18:35 +0000)]
Remove an unused variable.

15 years agoUpdate man pages after VFS_* changes in r191990.
tmclaugh [Sun, 24 May 2009 18:34:54 +0000 (18:34 +0000)]
Update man pages after VFS_* changes in r191990.

Approved by: brueffer, attilio

15 years agoFix typo in the manual page.
trasz [Sun, 24 May 2009 17:08:00 +0000 (17:08 +0000)]
Fix typo in the manual page.

15 years agoFix comment.
trasz [Sun, 24 May 2009 15:48:48 +0000 (15:48 +0000)]
Fix comment.

15 years agoFix elapsed (etime) field for swapped out processes in ps:
jilles [Sun, 24 May 2009 15:32:34 +0000 (15:32 +0000)]
Fix elapsed (etime) field for swapped out processes in ps:
show '-' instead of time since the Epoch.

PR: bin/123069
Submitted by: Vladimir Kozbin
Approved by: ed (mentor)
MFC after: 3 weeks

15 years agoFix chflags -h in various cases: do not use link target's flags as original.
jilles [Sun, 24 May 2009 15:27:25 +0000 (15:27 +0000)]
Fix chflags -h in various cases: do not use link target's flags as original.

Patch slightly changed to align more with chmod.c.

PR: bin/131999
Submitted by: bde
Approved by: ed (mentor)
MFC after: 3 weeks

15 years agoMake sure we feed 32bit align memory to nfsm_dissect otherwise we will fault
dfr [Sun, 24 May 2009 13:22:00 +0000 (13:22 +0000)]
Make sure we feed 32bit align memory to nfsm_dissect otherwise we will fault
on platforms with strict alignment requirements. In particular, this fixes the
problems with the new RPC transport on the arm platform.

Note: this adds yet another copy of nfs_realign(). I will attempt to refactor
after NFS_LEGACYRPC is removed.

Submitted by: sam

15 years agoThe advisory lock may be activated or activated and removed during the
kib [Sun, 24 May 2009 12:39:38 +0000 (12:39 +0000)]
The advisory lock may be activated or activated and removed during the
sleep waiting for conditions when the lock may be granted.
To prevent lf_setlock() from accessing possibly freed memory, add reference
counting to the struct lockf_entry. Bump refcount around the sleep.

Make lf_free_lock() return non-zero when structure was freed, and use
this after the sleep to return EINTR to the caller. The error code might
need a clarification, but we cannot return success to usermode, since
the lock is not owned anymore.

Reviewed by: dfr
Tested by: pho
MFC after: 1 month

15 years agoIn lf_purgelocks(), assert that state->ls_pending is empty after we
kib [Sun, 24 May 2009 12:37:55 +0000 (12:37 +0000)]
In lf_purgelocks(), assert that state->ls_pending is empty after we
weeded out threads, and clean ls_active instead of ls_pending.

Reviewed by: dfr
Tested by: pho
MFC after: 1 month

15 years agoIn lf_advlockasync(), recheck for doomed vnode after the state->ls_lock
kib [Sun, 24 May 2009 12:33:16 +0000 (12:33 +0000)]
In lf_advlockasync(), recheck for doomed vnode after the state->ls_lock
is acquired. In the lf_purgelocks(), assert that vnode is doomed and set
*statep to NULL before clearing ls_pending list. Otherwise, we allow for
the thread executing lf_advlockasync() to put new pending entry after
state->ls_lock is dropped in lf_purgelocks().

Reviewed by: dfr
Tested by: pho
MFC after: 1 month

15 years agoBlock when initially opening a TTY multiple times.
ed [Sun, 24 May 2009 12:32:03 +0000 (12:32 +0000)]
Block when initially opening a TTY multiple times.

In the original MPSAFE TTY code, I changed the behaviour by returning
EBUSY. I thought this made more sense, because it's basically a race to
see who gets the TTY first.

It turns out this is not a good change, because it also causes EBUSY to
be returned when another process is closing the TTY. This can happen
during startup, when /etc/rc (or one of its children) is still busy
draining its data and /sbin/init is attempting to open the TTY to spawn
a getty.

Reported by: bz
Tested by: bz

15 years agoReplace the while statement with the if for clarity. The loop body
kib [Sun, 24 May 2009 12:28:38 +0000 (12:28 +0000)]
Replace the while statement with the if for clarity. The loop body
cannot be executed more then once.

Reviewed by: dfr
Tested by: pho
MFC after: 1 month

15 years agoIncrease the number of available file descriptors to 64. This fixes the
dfr [Sun, 24 May 2009 11:37:10 +0000 (11:37 +0000)]
Increase the number of available file descriptors to 64. This fixes the
reported zfsboot problems for systems where more than seven drives are part
of ZFS pools.

15 years agoFix build of KGSSAPI bits post-vimage.
dfr [Sun, 24 May 2009 11:10:27 +0000 (11:10 +0000)]
Fix build of KGSSAPI bits post-vimage.

15 years agoAdd new constants to the acl(9) manual page.
trasz [Sun, 24 May 2009 09:42:53 +0000 (09:42 +0000)]
Add new constants to the acl(9) manual page.