]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years agouse a const format string in the log message so we can check the
luigi [Mon, 16 Feb 2009 12:09:52 +0000 (12:09 +0000)]
use a const format string in the log message so we can check the
arguments (if/when we enable those checks)

15 years agoIFF_NEEDSGIANT will no longer be supported, so remove compatibility code
rwatson [Mon, 16 Feb 2009 10:29:03 +0000 (10:29 +0000)]
IFF_NEEDSGIANT will no longer be supported, so remove compatibility code
from if_sppp framework for interfaces requiring Giant.

15 years agoAdd support to add, delete and modify logical partitions, as well
marcel [Mon, 16 Feb 2009 03:54:28 +0000 (03:54 +0000)]
Add support to add, delete and modify logical partitions, as well
as to create and destroy the extended partitioning scheme. In
other words: full support.

15 years agoAdd a helper function for loading geli keys from the loader.
thompsa [Mon, 16 Feb 2009 02:42:17 +0000 (02:42 +0000)]
Add a helper function for loading geli keys from the loader.

15 years agoAdd uslcom to the build too.
thompsa [Sun, 15 Feb 2009 23:40:29 +0000 (23:40 +0000)]
Add uslcom to the build too.

Reminded by: Michael Butler

15 years agoMake uslcom compile, cut'n'paste errors from uplcom.
thompsa [Sun, 15 Feb 2009 23:38:58 +0000 (23:38 +0000)]
Make uslcom compile, cut'n'paste errors from uplcom.

15 years agoRemove debug.ce.mpsafenet: we no longer support running the network
rwatson [Sun, 15 Feb 2009 23:29:13 +0000 (23:29 +0000)]
Remove debug.ce.mpsafenet: we no longer support running the network
stack with conditional Giant acquisition, and IFF_NEEDSGIANT will
be removed in the near future.

15 years agoRemove debug.ctau.mpsafenet: we no longer support running the network
rwatson [Sun, 15 Feb 2009 23:21:52 +0000 (23:21 +0000)]
Remove debug.ctau.mpsafenet: we no longer support running the network
stack with conditional Giant acquisition, and IFF_NEEDSGIANT will
be removed in the near future.

15 years agoRemove debug.cp.mpsafenet: we no longer support running the network
rwatson [Sun, 15 Feb 2009 23:18:29 +0000 (23:18 +0000)]
Remove debug.cp.mpsafenet: we no longer support running the network
stack with conditional Giant acquisition, and IFF_NEEDSGIANT will
be removed in the near future.

15 years agoSwitch over GENERIC kernels to USB2 by default.
thompsa [Sun, 15 Feb 2009 22:33:44 +0000 (22:33 +0000)]
Switch over GENERIC kernels to USB2 by default.

Tested by: make universe

15 years agoAdd method precheck to the g_part interface. The precheck
marcel [Sun, 15 Feb 2009 22:18:16 +0000 (22:18 +0000)]
Add method precheck to the g_part interface. The precheck
method allows schemes to reject the ctl request, pre-check
the parameters and/or modify/set parameters. There are 2
use cases that triggered the addition:
1.  When implementing a R/O scheme, deletes will still
    happen to the in-memory representation. The scheme is
    not involved in that operation. The pre-check method
    can be used to fail the delete up-front. Without this
    the write to disk will typically fail, but at that
    time the delete already happened.
2.  The EBR scheme uses a linked list to record slices.
    There's no index. The EBR scheme defines the index
    as a function of the start LBA of the partition. The
    add verb picks an index for the range and then invokes
    the add method of the scheme to fill in the blanks. It
    is too late for the add method to change the index.
    The pre-check is used to set the index up-front. This
    also (silently) overrides/nullifies any (pointless)
    user-specified index value.

15 years agoRemove unneeded AND operation. We have already checked that bit is set.
mav [Sun, 15 Feb 2009 21:54:16 +0000 (21:54 +0000)]
Remove unneeded AND operation. We have already checked that bit is set.

15 years agoEnable MSI support by default. Still can be disabled using device.hints.
mav [Sun, 15 Feb 2009 21:17:21 +0000 (21:17 +0000)]
Enable MSI support by default. Still can be disabled using device.hints.

15 years agoAdd initial single-vector MSI support into atapci driver.
mav [Sun, 15 Feb 2009 20:37:55 +0000 (20:37 +0000)]
Add initial single-vector MSI support into atapci driver.
Works fine with AHCI and theoretically other MSI capable devices.

At this moment support disabled by default. To enable it, set
"hint.atapci.X.msi=1" device hint.

15 years agoPull in kbd.c with usb2_input_kbd, just like ukbd.
thompsa [Sun, 15 Feb 2009 20:24:21 +0000 (20:24 +0000)]
Pull in kbd.c with usb2_input_kbd, just like ukbd.

15 years agoIt appears that none of the contents of this file are necessary, so
imp [Sun, 15 Feb 2009 20:05:13 +0000 (20:05 +0000)]
It appears that none of the contents of this file are necessary, so
replace the amd64-ish version with a blank version.

15 years agoAdd the pccard header files to ObsoleteFiles.inc.
ed [Sun, 15 Feb 2009 18:30:44 +0000 (18:30 +0000)]
Add the pccard header files to ObsoleteFiles.inc.

Reviewed by: imp

15 years agoRemoval pccard directory requires removing it from the list of things
imp [Sun, 15 Feb 2009 18:19:24 +0000 (18:19 +0000)]
Removal pccard directory requires removing it from the list of things
that get installed too.  <blush>

Submitted by: Olivier Smedts (identical patches to what I came up
with before checking email)

15 years agoMake dumper_t definition conform more closely to stlye(9). This also
imp [Sun, 15 Feb 2009 17:56:16 +0000 (17:56 +0000)]
Make dumper_t definition conform more closely to stlye(9).  This also
avoid keywords in other languages that may have been present before.

Submitted by: Andriy Gapon, jkoshy@

15 years agoNow that ifmcstat(8) does not suck, retire host-mode netstat -g.
bms [Sun, 15 Feb 2009 16:16:38 +0000 (16:16 +0000)]
Now that ifmcstat(8) does not suck, retire host-mode netstat -g.
This change will not be back-ported.

15 years agoAdd workaround for some ATI chips, failing to soft-reset
mav [Sun, 15 Feb 2009 16:05:48 +0000 (16:05 +0000)]
Add workaround for some ATI chips, failing to soft-reset
when port multiplicator supported, but absent.

15 years agonits
bms [Sun, 15 Feb 2009 15:26:42 +0000 (15:26 +0000)]
nits

15 years agoBlow away KAME MLDv2 hooks.
bms [Sun, 15 Feb 2009 15:21:34 +0000 (15:21 +0000)]
Blow away KAME MLDv2 hooks.
This code can be gotten from change history here and it's
more than likely our implementation will differ significantly
because of VIMAGE and SMPng.

15 years agoImprove ifmcstat(8) and fix a few bugs while we're at it:
bms [Sun, 15 Feb 2009 15:19:34 +0000 (15:19 +0000)]
Improve ifmcstat(8) and fix a few bugs while we're at it:
 * Retire the old 'ifmcstat <kernel>' usage.
 * Print AF_LINK records even if run against KVM.
   This makes the KVM backend consistent with the sysctl backend.
 * Suppress printing of link-layer group records by default.
 * Add a -v switch to allow link-layer groups to be printed.
 * If compiled without INET6 support, actually work.
 * If compiled with INET6 support, print the scope ID of
   all IPv6 addresses in both backends.
 * Update man page.
 * Update copyrights.

With this change, it is now reasonable to retire netstat -g.
Most of the SSM related gunk in this file will require later refactoring.

MFC after: 2 weeks

15 years ago- Do not free the pattern lists immediately after use, as they might be needed
lulf [Sun, 15 Feb 2009 13:22:21 +0000 (13:22 +0000)]
- Do not free the pattern lists immediately after use, as they might be needed
  again in case the connection is interrupted and csup have to reconnect. The
  lists will be freed after the collection has been completely processed.

PR: bin/131477
Tested by: dchagin

15 years agoFix a typo which caused ifmcstat's sysctl path
bms [Sun, 15 Feb 2009 12:10:05 +0000 (12:10 +0000)]
Fix a typo which caused ifmcstat's sysctl path
to print the network-layer endpoint address of the
group membership, rather than its link-layer mapping
as intended.
The KVM path is not affected.

MFC after: 1 week

15 years agosys/pccard is gone.
nyan [Sun, 15 Feb 2009 11:05:50 +0000 (11:05 +0000)]
sys/pccard is gone.

15 years agoTune output to remove trailing space.
mav [Sun, 15 Feb 2009 10:41:42 +0000 (10:41 +0000)]
Tune output to remove trailing space.

Submitted by: Christoph Mallon

15 years agoAdd SATA and PCI Advanced Features capabilities reporting.
mav [Sun, 15 Feb 2009 09:56:47 +0000 (09:56 +0000)]
Add SATA and PCI Advanced Features capabilities reporting.

15 years agoAdd SATA and PCI Advanced Features capabilities constants.
mav [Sun, 15 Feb 2009 09:49:21 +0000 (09:49 +0000)]
Add SATA and PCI Advanced Features capabilities constants.

15 years ago- fix formatting
kmacy [Sun, 15 Feb 2009 06:36:02 +0000 (06:36 +0000)]
- fix formatting
- fix types in ticks_to_system_time

15 years agoActually include geom_part_ebr.c when options GEOM_PART_EBR is
marcel [Sun, 15 Feb 2009 03:50:27 +0000 (03:50 +0000)]
Actually include geom_part_ebr.c when options GEOM_PART_EBR is
present.

Pointy hat: marcel

15 years agoDon't remove gdbserver on ARM. It's not obsolete. In fact,
marcel [Sun, 15 Feb 2009 03:42:14 +0000 (03:42 +0000)]
Don't remove gdbserver on ARM. It's not obsolete. In fact,
it's quite nouveau...

15 years agoThis directory is no logner needed after copy to usr.sbin/dumpcis.
imp [Sun, 15 Feb 2009 03:22:15 +0000 (03:22 +0000)]
This directory is no logner needed after copy to usr.sbin/dumpcis.

15 years agoMove sys/pccard/*.h here.
imp [Sun, 15 Feb 2009 03:21:37 +0000 (03:21 +0000)]
Move sys/pccard/*.h here.

15 years agoRemove unnecessary pccard/cardinfo.h include.
imp [Sun, 15 Feb 2009 03:12:46 +0000 (03:12 +0000)]
Remove unnecessary pccard/cardinfo.h include.

15 years agoRemove vestiges of OLDCARD PC Card support. We haven't needed/used
imp [Sun, 15 Feb 2009 03:10:20 +0000 (03:10 +0000)]
Remove vestiges of OLDCARD PC Card support.  We haven't needed/used
this since 5.x.

15 years agoThis directory isn't needed anymore.
imp [Sun, 15 Feb 2009 03:04:20 +0000 (03:04 +0000)]
This directory isn't needed anymore.

15 years agoTruss builds on mips, so don't filter it building there anymore.
imp [Sun, 15 Feb 2009 01:32:50 +0000 (01:32 +0000)]
Truss builds on mips, so don't filter it building there anymore.

15 years agoAdd preliminary support for truss on MIPS. It compiles, but has not
imp [Sun, 15 Feb 2009 01:26:49 +0000 (01:26 +0000)]
Add preliminary support for truss on MIPS.  It compiles, but has not
been extensively tested.  And the ELF64 stuff likely is not quite
right...

# There's a lot of cut-n-paste code here that could easily be
# refactored, at least for FreeBSD syscalls.

15 years agoRemove stray __P()
imp [Sun, 15 Feb 2009 01:12:16 +0000 (01:12 +0000)]
Remove stray __P()

15 years agoremove unnecessary #include from vnet.h and vinet.h
luigi [Sun, 15 Feb 2009 00:28:28 +0000 (00:28 +0000)]
remove unnecessary #include from vnet.h and vinet.h

Approved by: Marko Zec

15 years agoMFp4 //depot/projects/usb@157699
thompsa [Sat, 14 Feb 2009 23:20:00 +0000 (23:20 +0000)]
MFp4 //depot/projects/usb@157699

Add two new functions to the libusb20 API and required kernel ioctls.

- libusb20_dev_get_iface_desc
- libusb20_dev_get_info

New command to usbconfig, "show_ifdrv", which will print out the kernel driver
attached to the given USB device aswell.

See "man libusb20" for a detailed description.

Some minor style corrections long-line wrapping.

Submitted by: Hans Petter Selasky

15 years agoTunes to AHCI reset sequences:
mav [Sat, 14 Feb 2009 23:02:59 +0000 (23:02 +0000)]
Tunes to AHCI reset sequences:
- specification claims that 1 second is just a maximum controller reset time;
implement controller reset properly to save almost 1 second of boot, and
about half second of resume time;
- enable channel interrupts only after channel status reset to fix duplicate
device creation on resume due to unwanted device connection event;
- as described in specification, wait for disk ready status after channel
power-up; it is not so important when disk already touched by BIOS, but
solves device not ready problems on resume and probably some other cases.
- uncomment channel stop/start on soft-reset as it is declared mandatory by
specification; it was commented due to some random drive detection problems
on VIA and JMicron controllers, but I hope it is fixed by previous point.

15 years ago- ieee80211_chan2ieee returns an int
thompsa [Sat, 14 Feb 2009 22:29:54 +0000 (22:29 +0000)]
- ieee80211_chan2ieee returns an int
- avoid null deref in detach
- update pause timings

Obtained from: //depot/projects/usb

15 years agoRemove unnecessary page queues locking around vm_page_wakeup().
alc [Sat, 14 Feb 2009 22:07:22 +0000 (22:07 +0000)]
Remove unnecessary page queues locking around vm_page_wakeup().

Approved by: kmacy

15 years agoRemove the printf's when the vnode to be exported for procstat is not a VDIR.
marcus [Sat, 14 Feb 2009 21:55:09 +0000 (21:55 +0000)]
Remove the printf's when the vnode to be exported for procstat is not a VDIR.
If the file system backing a process' cwd is removed, and procstat -f PID
is called, then these messages would have been printed.  The extra verbosity is
not required in this situation.

Requested by: kib
Approved by: kib

15 years agoDEVICE_PROBE(9) claims that we must not initialize softc on probe stage.
mav [Sat, 14 Feb 2009 21:54:44 +0000 (21:54 +0000)]
DEVICE_PROBE(9) claims that we must not initialize softc on probe stage.
Move channel softc initialization from ata_XXX_probe() to ata_XXX_attach().

Instead of calculating ata channel number as position in child device list,
pass it's real number directly from controller probe routine using ivars.
It is simpler and IMHO more correct.

15 years agoChange two KASSERTS to printfs and simple returns. Stress testing has
marcus [Sat, 14 Feb 2009 21:12:24 +0000 (21:12 +0000)]
Change two KASSERTS to printfs and simple returns.  Stress testing has
revealed that a process' current working directory can be VBAD if the
directory is removed.  This can trigger a panic when procstat -f PID is
run.

Tested by: pho
Discovered by: phobot
Reviewed by: kib
Approved by: kib

15 years agoRemove unneeded execute-permissions from the USB2 device node.
ed [Sat, 14 Feb 2009 19:46:28 +0000 (19:46 +0000)]
Remove unneeded execute-permissions from the USB2 device node.

Giving a charactere device execute permissions doesn't have any use.
Right now there isn't a single device node in /dev that has it, except
the USB2 device node, so remove it.

Approved by: hps, thompsa

15 years agoRemove unnecessary page queues locking around vm_page_busy() and
alc [Sat, 14 Feb 2009 18:23:52 +0000 (18:23 +0000)]
Remove unnecessary page queues locking around vm_page_busy() and
vm_page_wakeup().  (This change is applicable to RELENG_7 but not
RELENG_6.)

MFC after: 1 week

15 years agoRestore MAKE_FLOPPIES for pc98.
nyan [Sat, 14 Feb 2009 14:53:51 +0000 (14:53 +0000)]
Restore MAKE_FLOPPIES for pc98.

15 years agoSpecify 4096 block-size and 512 fragment-size.
nyan [Sat, 14 Feb 2009 14:50:27 +0000 (14:50 +0000)]
Specify 4096 block-size and 512 fragment-size.
This fixes no space error.

15 years agoThis commit fixes the issue with alias_sctp.c. No
rrs [Sat, 14 Feb 2009 11:34:57 +0000 (11:34 +0000)]
This commit fixes the issue with alias_sctp.c. No
longer do we require SCTP to be in the kernel for the
lib to be able to handle SCTP. We do this by moving
the CRC32c checksum into libkern/crc32.c and then adjusting
all routines to use the common methods. Note that this
will improve the performance of iSCSI since they were
using the old single 256 bit table lookup versus the
slicing 8 algorithm (which gives a 4x speed up in
CRC32c calculation :-D)

Reviewed by:rwatson, gnn, scottl, paolo
MFC after: 4 week? (assuming we MFC the alias_sctp changes)

15 years agoSince r188030 the error value for attach is returned, this means if kbdmux
thompsa [Fri, 13 Feb 2009 23:36:08 +0000 (23:36 +0000)]
Since r188030 the error value for attach is returned, this means if kbdmux
fails to attach (possibly due to disable hints) then we get called back for
unload. Correctly handle the case where the keyboard isnt found rather than
calling panic.

15 years agoSync comment with actual configuration format.
delphij [Fri, 13 Feb 2009 22:48:05 +0000 (22:48 +0000)]
Sync comment with actual configuration format.

15 years ago- ieee80211_chan2ieee returns an int
thompsa [Fri, 13 Feb 2009 21:45:19 +0000 (21:45 +0000)]
- ieee80211_chan2ieee returns an int
- set ic_update_promisc to the same callback as mcast
- avoid null deref in zyd_detach

Obtained from: //depot/projects/usb

15 years agoMFp4 //depot/projects/usb; 157501, 157608, 157609
thompsa [Fri, 13 Feb 2009 20:57:43 +0000 (20:57 +0000)]
MFp4 //depot/projects/usb; 157501, 157608, 157609

 - Make usb2_transfer_pending() part of the USB core header file.
 - Make usb2_transfer_pending() NULL safe.
 - Make sure that USB process functions return if the process has been drained.
 - Remove two unused functions.

Submitted by:   Hans Petter Selasky

15 years agoElaborate some on the workings of the stub.
n_hibma [Fri, 13 Feb 2009 20:09:11 +0000 (20:09 +0000)]
Elaborate some on the workings of the stub.
Collapse up the list of supported devices.

15 years agoAdd support for CMOTECH devices (not sure whether this is the correct
n_hibma [Fri, 13 Feb 2009 19:49:51 +0000 (19:49 +0000)]
Add support for CMOTECH devices (not sure whether this is the correct
name) (not sure whether this works correctly, but should be close).

Fix the stub attach phase for some Novatel cards. They expect the CSW
(repsonse to CBW, SCSI eject command) to be fetched before switching to
modem mode.

MFC after: 2 weeks

15 years agobridge_delete_member is called via the event handler from if_detach
thompsa [Fri, 13 Feb 2009 19:20:25 +0000 (19:20 +0000)]
bridge_delete_member is called via the event handler from if_detach
after the LLADDR is reclaimed which causes a null pointer deref with
inherit_mac enabled. Record the ifnet pointer of the interface and then compare
that to find when to re-assign the bridge address.

Submitted by: sam

15 years agoRemove semicolon left in the last commit
thompsa [Fri, 13 Feb 2009 18:51:39 +0000 (18:51 +0000)]
Remove semicolon left in the last commit

Spotted by: csjp

15 years agoAdd the Novatel U760.
thompsa [Fri, 13 Feb 2009 18:45:36 +0000 (18:45 +0000)]
Add the Novatel U760.

15 years agoHave the jail code use the error returned to pass not constant
rrs [Fri, 13 Feb 2009 18:44:30 +0000 (18:44 +0000)]
Have the jail code use the error returned to pass not constant
errors.
Obtained from: jamie@freebsd.org

15 years agoUse shared vnode locks when invoking VOP_READDIR().
jhb [Fri, 13 Feb 2009 18:18:14 +0000 (18:18 +0000)]
Use shared vnode locks when invoking VOP_READDIR().

MFC after: 1 month

15 years agoUpdate the Self ID structure to conform to 1394a-2000.
sbruno [Fri, 13 Feb 2009 17:45:09 +0000 (17:45 +0000)]
Update the Self ID structure to conform to 1394a-2000.

Delete the unused defines.

Reviewed by: scottl

15 years agoRemove redundant while () from loop.
sbruno [Fri, 13 Feb 2009 17:44:07 +0000 (17:44 +0000)]
Remove redundant while () from loop.

Submitted by:  Ganbold <ganbold@micom.mng.net>
Reviewed by:  scottl

15 years agoHonor WITHOUT_INSTALLLIB in some places.
jkim [Fri, 13 Feb 2009 16:51:36 +0000 (16:51 +0000)]
Honor WITHOUT_INSTALLLIB in some places.

15 years agoremove references to ic_stats
sam [Fri, 13 Feb 2009 16:17:05 +0000 (16:17 +0000)]
remove references to ic_stats

Spotted by: Lucius Windschuh <lwindschuh@googlemail.com>

15 years agoOops. ARM_RAS_END is ARM_TP_ADDRESS + 8, not 4.
cognet [Fri, 13 Feb 2009 16:00:19 +0000 (16:00 +0000)]
Oops. ARM_RAS_END is ARM_TP_ADDRESS + 8, not 4.

Spotted out by: Mark Tinguely <tinguely at casselton d0t net>

15 years agoremove unnecessary #include, and document some of the others
luigi [Fri, 13 Feb 2009 15:37:14 +0000 (15:37 +0000)]
remove unnecessary #include, and document some of the others

15 years agoFix a bug in the previous change to the mtab handler: use the path returned
jhb [Fri, 13 Feb 2009 15:32:03 +0000 (15:32 +0000)]
Fix a bug in the previous change to the mtab handler: use the path returned
by vn_fullpath() when vn_fullpath() succeeds instead of when it fails.

Submitted by: Artem Belevich  fbsdlist of src.cx
MFC after: 3 days

15 years agoUse uint32_t instead of n_long and n_time, and uint16_t instead of n_short.
luigi [Fri, 13 Feb 2009 15:14:43 +0000 (15:14 +0000)]
Use uint32_t instead of n_long and n_time, and uint16_t instead of n_short.
Add a note next to fields in network format.

The n_* types are not enough for compiler checks on endianness, and their
use often requires an otherwise unnecessary #include <netinet/in_systm.h>

The typedef in in_systm.h are still there.

15 years agoMove the new rwnd field down to the very end
rrs [Fri, 13 Feb 2009 14:43:46 +0000 (14:43 +0000)]
Move the new rwnd field down to the very end
of the xsctp structure. This is where all new
fields belong (not that we will be ABI compatiable
with 7.x anyway.. sigh).

15 years agoo In case of the error do not forget to deallocate a cloned device unit.
maxim [Fri, 13 Feb 2009 12:59:54 +0000 (12:59 +0000)]
o In case of the error do not forget to deallocate a cloned device unit.

PR: kern/131642
Submitted by: Dmitrij Tejblum
MFC after: 1 week

15 years agoFix an edge-case of the linux readdir: We need the size of a linux dirent
netchild [Fri, 13 Feb 2009 11:55:19 +0000 (11:55 +0000)]
Fix an edge-case of the linux readdir: We need the size of a linux dirent
structure, not the size of a pointer to it.

PR: 131099
Submitted by: Andreas Kies <andikies@gmail.com>
MFC after: 2 weeks

15 years agoClarify and reimplement the bioq API so that bioq_disksort() has
luigi [Fri, 13 Feb 2009 11:36:32 +0000 (11:36 +0000)]
Clarify and reimplement the bioq API so that bioq_disksort() has
the correct behaviour (sorting by distance from the current head position
in the scan direction) and bioq_insert_head() and bioq_insert_tail()
have a well defined (and useful) behaviour, especially when intermixed
with calls to bioq_disksort().

In particular:
- fix a bug in the existing bioq_disksort() that did not use the
  current head position correctly;
- redefine semantics of bioq_insert_head() and bioq_insert_tail().
  bioq_insert_tail() can now be used as a barrier
  between previous and subsequent calls to bioq_disksort().

The code is heavily documented in the source code so please refer
to that for the details.

Much of this code comes from Fabio Checconi. Also thanks to Kirk
for feedback on the (re)definition of bioq_insert_tail().

NOTE: in the current tree there is only a handful of files which
intermix calls to bioq_disksort() with bioq_insert_head() and
bioq_insert_tail(). The ordering of the queue in these situation
was not specified (nor easy to figure out) before, so I doubt any
of that code could be affected by the specification of the API.

Also note that the current implementation is significantly simpler
than the previous one (also used in ata_sort_queue()).
It would be useful to reimplement ata_sort_queue() using
the same code used in bioq_disksort().

MFC after: 1 week

15 years agoIn the case that the probe has determined that it can't query the device for
scottl [Fri, 13 Feb 2009 10:04:59 +0000 (10:04 +0000)]
In the case that the probe has determined that it can't query the device for
a serial number, fall through to the next case so that initial negotiation
still happens.  Without this, devices were showing up with only 1 available
tag opening, leading to observations of very poor I/O performance.

This should fix problems reported with VMWare Fusion and ESX.  Early
generation MPT-SAS controllers with SATA disks might also be affected.
HP CISS controllers are also likely affected, as are many other
pseudo-scsi disk subsystems.

15 years agoadd -z option to zero driver statistics (but note we don't yet zero
sam [Fri, 13 Feb 2009 05:45:23 +0000 (05:45 +0000)]
add -z option to zero driver statistics (but note we don't yet zero
stats obtained from the hal)

15 years agoadd stats no frags and missing stat for no dma buffers
sam [Fri, 13 Feb 2009 05:44:18 +0000 (05:44 +0000)]
add stats no frags and missing stat for no dma buffers

15 years agobackout r188327, it broke summary printing
sam [Fri, 13 Feb 2009 05:42:25 +0000 (05:42 +0000)]
backout r188327, it broke summary printing

15 years agoadd SIOCZATHSTATS ioctl to zero driver statistics
sam [Fri, 13 Feb 2009 05:38:03 +0000 (05:38 +0000)]
add SIOCZATHSTATS ioctl to zero driver statistics

15 years agoremove ic_stats; it was intended to accumulate stats from vaps as they
sam [Fri, 13 Feb 2009 05:31:18 +0000 (05:31 +0000)]
remove ic_stats; it was intended to accumulate stats from vaps as they
were reaped but was never used and is inaccessible

15 years agoadd driver stat to count tx drops due to insufficient frag buffers
sam [Fri, 13 Feb 2009 05:30:22 +0000 (05:30 +0000)]
add driver stat to count tx drops due to insufficient frag buffers

15 years agostrip M_FRAG & co mbufs coming down from ip_output, these flags are
sam [Fri, 13 Feb 2009 05:26:04 +0000 (05:26 +0000)]
strip M_FRAG & co mbufs coming down from ip_output, these flags are
re-used within net80211 to mark 802.11 frags so allowing them to
leak through to the driver caused packets to be dropped in ath

15 years agoRestore some of the link handling code that was inadvertently removed in
thompsa [Fri, 13 Feb 2009 03:38:08 +0000 (03:38 +0000)]
Restore some of the link handling code that was inadvertently removed in
r188412, this broke the GU-1000T adapters.

Submitted by: yongari
Pointy hat: me

15 years agoAdd module dependency on miibus.
thompsa [Fri, 13 Feb 2009 02:48:32 +0000 (02:48 +0000)]
Add module dependency on miibus.

Submitted by: yongari

15 years agoReceving VLAN oversized frames raise SIS_ISR_RX_ERR interrupt, so
yongari [Fri, 13 Feb 2009 02:08:20 +0000 (02:08 +0000)]
Receving VLAN oversized frames raise SIS_ISR_RX_ERR interrupt, so
make sis_rxeof() handle that too. Previously it used to receive the
frame and reset controller.

PR: kern/131414

15 years agoRecognize AR5212_AR2317_REV2 in ar5312Probe()
sam [Fri, 13 Feb 2009 01:50:04 +0000 (01:50 +0000)]
Recognize AR5212_AR2317_REV2 in ar5312Probe()

Submitted by: Pavel Roskin <proski@gnu.org>

15 years agoCheck the exit flag at the start of the taskqueue loop rather than the end. It
thompsa [Fri, 13 Feb 2009 01:16:51 +0000 (01:16 +0000)]
Check the exit flag at the start of the taskqueue loop rather than the end. It
is possible to tear down the taskqueue before the thread has run and the
taskqueue loop would sleep forever.

Reviewed by: sam
MFC after: 1 week

15 years agoRemove unused ifaddr local variable in ioctl routine.
rwatson [Fri, 13 Feb 2009 00:01:11 +0000 (00:01 +0000)]
Remove unused ifaddr local variable in ioctl routine.

MFC after: 3 days

15 years agoRemove unused ifaddr and ifreq local variables.
rwatson [Thu, 12 Feb 2009 23:55:35 +0000 (23:55 +0000)]
Remove unused ifaddr and ifreq local variables.

MFC after: 3 days

15 years agoo replace ieee80211_node_reclaim by individual operations to release the
sam [Thu, 12 Feb 2009 23:34:58 +0000 (23:34 +0000)]
o replace ieee80211_node_reclaim by individual operations to release the
  references to iv_bss and the sta table; this is equivalent and causes
  direct reclaim of the old bss node when any references in packets inflight
  are reclaimed (previously the old node would sit in the bss table until
  the inactivity processing reclaimed it)
o remove ieee80211_node_reclaim now that it's only use is gone

Reviewed by: avatar, cbzimmer

15 years agoTo prevent various race conditions in the RAS code, store and restore the
cognet [Thu, 12 Feb 2009 23:23:30 +0000 (23:23 +0000)]
To prevent various race conditions in the RAS code, store and restore the
values in ARM_RAS_START and ARM_RAS_END at context switch time.

MFC after: 1 week

15 years agoDo not set thread0.td_frame to a bogus value, as it's going to overwrite the
cognet [Thu, 12 Feb 2009 22:55:39 +0000 (22:55 +0000)]
Do not set thread0.td_frame to a bogus value, as it's going to overwrite the
thread0 pcb, while the board-dependant code already set a good trapframe.

Reported by: Mark Tinguely <tinguely at casselton d0t net>

MFC after: 1 week

15 years agoUnbreak build by making svr4 conditional on i386.
cognet [Thu, 12 Feb 2009 21:01:16 +0000 (21:01 +0000)]
Unbreak build by making svr4 conditional on i386.

15 years agoProperly invalidate highest pts number when calling setttyent().
ed [Thu, 12 Feb 2009 19:32:52 +0000 (19:32 +0000)]
Properly invalidate highest pts number when calling setttyent().

When calling setttyent() after calling endttyent(), pts_valid will never
be set to 1, because the readdir()-loop will likely never vind a pts
that has a higher number than before.

Simplify the code by removing pts_valid. We'll just set maxpts to -1
when we don't have a valid count yet.

15 years agoRemove pts(4) entries from /etc/ttys.
ed [Thu, 12 Feb 2009 19:21:48 +0000 (19:21 +0000)]
Remove pts(4) entries from /etc/ttys.

Even though I increased the amount of pts(4) entries in /etc/ttys some
time ago, I didn't realize back then those entries shouldn't have been
there in the first place.

I just looked at the getttyent() source code and it turns out when you
call setttyent(), it walks through /dev/pts and looks for the device
with the highest number. After you receive EOF's from getttyent(), it
makes up entries for pts(4) devices.

This means that adding entries for pts(4) is somewhat harmful, because
if you now traverse the list, you get redundant entries, so just remove
them.

15 years agoMake ttyslot(3) work with pts(4) devices.
ed [Thu, 12 Feb 2009 19:00:13 +0000 (19:00 +0000)]
Make ttyslot(3) work with pts(4) devices.

It seems ttyslot() calls rindex(), to strip the device name to the last
slash, but this is obviously invalid. /dev/pts/0 should be stripped
until pts/0. Because /etc/ttys only supports TTY names in /dev/, just
strip this piece of the pathname.

15 years agoAdd a ieee80211_waitfor_parent() function that will wait for all deferred
thompsa [Thu, 12 Feb 2009 18:57:18 +0000 (18:57 +0000)]
Add a ieee80211_waitfor_parent() function that will wait for all deferred
parent interface tasks to complete. This had been added to the ioctl path but
it is also need elsewhere like detach so its safe to teardown.

Reported by: Hans Petter Selasky
Submitted by: sam