]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoMuch more correct EFTYPE description.
pjd [Wed, 13 Dec 2006 13:46:01 +0000 (13:46 +0000)]
Much more correct EFTYPE description.

17 years agoBetter wording.
pjd [Wed, 13 Dec 2006 13:26:32 +0000 (13:26 +0000)]
Better wording.

17 years agoAppend-only flag also denies chmod(2). Is this correct behaviour?
pjd [Wed, 13 Dec 2006 13:22:57 +0000 (13:22 +0000)]
Append-only flag also denies chmod(2). Is this correct behaviour?

17 years agoRefine the previous revision.
ru [Wed, 13 Dec 2006 12:09:41 +0000 (12:09 +0000)]
Refine the previous revision.

17 years ago"which -s" will return 0 only if all executables were found, not "any".
ru [Wed, 13 Dec 2006 12:07:49 +0000 (12:07 +0000)]
"which -s" will return 0 only if all executables were found, not "any".

17 years agoBe more precise with EPERM description. When chown(2) is a no-op, it will
pjd [Wed, 13 Dec 2006 11:46:38 +0000 (11:46 +0000)]
Be more precise with EPERM description. When chown(2) is a no-op, it will
return 0.

17 years agoWrite permission if of course only needed for the parent directory of
pjd [Wed, 13 Dec 2006 11:26:03 +0000 (11:26 +0000)]
Write permission if of course only needed for the parent directory of
the object beeing created.

Pointed out by: bde

17 years agomkfifo(2) returns EACCES when write permission is denied for a component of
pjd [Wed, 13 Dec 2006 09:58:49 +0000 (09:58 +0000)]
mkfifo(2) returns EACCES when write permission is denied for a component of
the path prefix.

17 years agoAdd Xref to msk(4).
brueffer [Wed, 13 Dec 2006 08:56:15 +0000 (08:56 +0000)]
Add Xref to msk(4).

17 years agoCorrect sort msk(4).
brueffer [Wed, 13 Dec 2006 08:55:20 +0000 (08:55 +0000)]
Correct sort msk(4).

17 years agoAutogenerate hardware notes for msk(4).
brueffer [Wed, 13 Dec 2006 08:53:27 +0000 (08:53 +0000)]
Autogenerate hardware notes for msk(4).

17 years agoMisc fixes, also sort the hardware list alphabetically and add more history.
brueffer [Wed, 13 Dec 2006 08:51:40 +0000 (08:51 +0000)]
Misc fixes, also sort the hardware list alphabetically and add more history.

17 years agoUpdate to version 1.20.00.13 of the arcmsr driver. This fixes many bugs
scottl [Wed, 13 Dec 2006 08:46:03 +0000 (08:46 +0000)]
Update to version 1.20.00.13 of the arcmsr driver.  This fixes many bugs
as well as adds support for newer controllers.

Many thanks to Areca for their continued support.

Submitted by: Erich Chen

17 years agoIt's ``apply echo *'' that is similar to ``ls -1'', not ``apply echo a*''.
ru [Wed, 13 Dec 2006 08:13:14 +0000 (08:13 +0000)]
It's ``apply echo *'' that is similar to ``ls -1'', not ``apply echo a*''.

Submitted by: sat

17 years ago- Fix prototype of bpf_filter().
ru [Wed, 13 Dec 2006 06:27:20 +0000 (06:27 +0000)]
- Fix prototype of bpf_filter().
- Document what passing it a zero buflen argument means.

17 years agoTeach login(1) about the make.conf NO_AUDIT variable. This allows us to
csjp [Wed, 13 Dec 2006 06:13:32 +0000 (06:13 +0000)]
Teach login(1) about the make.conf NO_AUDIT variable.  This allows us to
conditionally build in audit support.

Submitted by: bz
MFC after: 1 week

17 years agoImplement OF_decode_addr(). This makes uart(4) work as a serial
marcel [Wed, 13 Dec 2006 06:11:22 +0000 (06:11 +0000)]
Implement OF_decode_addr(). This makes uart(4) work as a serial
console on a Xserve G4.

17 years agoTeach the MAC policies which utilize mbuf labeling the new syncache
csjp [Wed, 13 Dec 2006 06:03:22 +0000 (06:03 +0000)]
Teach the MAC policies which utilize mbuf labeling the new syncache
entry points.  Properly initialize the mbuf label based on the label
we copy from the PCB. This fixes an LOR between the PCB and syncache
code.

17 years agoFix LOR between the syncache and inpcb locks when MAC is present in the
csjp [Wed, 13 Dec 2006 06:00:57 +0000 (06:00 +0000)]
Fix LOR between the syncache and inpcb locks when MAC is present in the
kernel.  This LOR snuck in with some of the recent syncache changes.  To
fix this, the inpcb handling was changed:

- Hang a MAC label off the syncache object
- When the syncache entry is initially created, we pickup the PCB lock
  is held because we extract information from it while initializing the
  syncache entry.  While we do this, copy the MAC label associated with
  the PCB and use it for the syncache entry.
- When the packet is transmitted, copy the label from the syncache entry
  to the mbuf so it can be processed by security policies which analyze
  mbuf labels.

This change required that the MAC framework be extended to support the
label copy operations from the PCB to the syncache entry, and then from
the syncache entry to the mbuf.

These functions really should be referencing the syncache structure instead
of the label.  However, due to some of the complexities associated with
exposing this syncache structure we operate directly on it's label pointer.
This should be OK since we aren't making any access control decisions within
this code directly, we are merely allocating and copying label storage so
we can properly initialize mbuf labels for any packets the syncache code
might create.

This also has a nice side effect of caching.  Prior to this change, the
PCB would be looked up/locked for each packet transmitted.  Now the label
is cached at the time the syncache entry is initialized.

Submitted by: andre [1]
Discussed with: rwatson

[1] andre submitted the tcp_syncache.c changes

17 years agoAdd msk(4) to the list of drivers supported by GENERIC kernel.
yongari [Wed, 13 Dec 2006 03:41:47 +0000 (03:41 +0000)]
Add msk(4) to the list of drivers supported by GENERIC kernel.

17 years agoImplement bus_space_map().
marcel [Wed, 13 Dec 2006 03:25:31 +0000 (03:25 +0000)]
Implement bus_space_map().

17 years agoAdd msk(4) to the list of supported network interface.
yongari [Wed, 13 Dec 2006 02:51:26 +0000 (02:51 +0000)]
Add msk(4) to the list of supported network interface.

17 years agoAdd an entry for the msk(4) module.
yongari [Wed, 13 Dec 2006 02:48:22 +0000 (02:48 +0000)]
Add an entry for the msk(4) module.

17 years agoAdd msk(4) to the list of device drivers use the miibus interface.
yongari [Wed, 13 Dec 2006 02:46:32 +0000 (02:46 +0000)]
Add msk(4) to the list of device drivers use the miibus interface.

17 years agoAdd msk(4) to the list of devices supporting full VLAN processing
yongari [Wed, 13 Dec 2006 02:42:01 +0000 (02:42 +0000)]
Add msk(4) to the list of devices supporting full VLAN processing
in the hardware.

17 years agoAdd msk(4) to the list of drivers supporting ALTQ.
yongari [Wed, 13 Dec 2006 02:40:06 +0000 (02:40 +0000)]
Add msk(4) to the list of drivers supporting ALTQ.

17 years agoHook up msk(4) to the build.
yongari [Wed, 13 Dec 2006 02:37:48 +0000 (02:37 +0000)]
Hook up msk(4) to the build.

17 years agoAdd a man page for msk(4).
yongari [Wed, 13 Dec 2006 02:37:01 +0000 (02:37 +0000)]
Add a man page for msk(4).

17 years agoAdd msk(4), a driver for Marvell/SysKonnect Yukon II Gigabit Ethernet
yongari [Wed, 13 Dec 2006 02:30:11 +0000 (02:30 +0000)]
Add msk(4), a driver for Marvell/SysKonnect Yukon II Gigabit Ethernet
controller. Due to lack of documentation, this driver is based on the
code from sk(4) and Marvell's myk(4) driver for FreeBSD. I've also
adopted the OpenBSD interface name, msk(4) in order to reduce naming
differences between BSDs.
The msk(4) driver supports the following Gigabit Ethernet adapters.

o SysKonnect SK-9Sxx Gigabit Ethernet
o SysKonnect SK-9Exx Gigabit Ethernet
o Marvell Yukon 88E8021CU Gigabit Ethernet
o Marvell Yukon 88E8021 SX/LX Gigabit Ethernet
o Marvell Yukon 88E8022CU Gigabit Ethernet
o Marvell Yukon 88E8022 SX/LX Gigabit Ethernet
o Marvell Yukon 88E8061CU Gigabit Ethernet
o Marvell Yukon 88E8061 SX/LX Gigabit Ethernet
o Marvell Yukon 88E8062CU Gigabit Ethernet
o Marvell Yukon 88E8062 SX/LX Gigabit Ethernet
o Marvell Yukon 88E8035 Gigabit Ethernet
o Marvell Yukon 88E8036 Gigabit Ethernet
o Marvell Yukon 88E8038 Gigabit Ethernet
o Marvell Yukon 88E8050 Gigabit Ethernet
o Marvell Yukon 88E8052 Gigabit Ethernet
o Marvell Yukon 88E8053 Gigabit Ethernet
o Marvell Yukon 88E8055 Gigabit Ethernet
o Marvell Yukon 88E8056 Gigabit Ethernet
o D-Link 550SX Gigabit Ethernet
o D-Link 560T Gigabit Ethernet

Unlike OpenBSD/NetBSD msk(4), the msk(4) driver supports all hardware
features including TCP/UDP checksum offload for transmit, MSI, TCP
segmentation offload(TSO), hardware VLAN tag stripping/insertion,
and jumbo frames(up to 9022 bytes). The only unsupported hardware
feature except RLMT is Rx checksum offload which I don't know how to
make it work reliably.

Known Issues:
 It seems msk(4) does not work on the second port of dual port NIC.
 (The first port works without problems.)

Thanks to Marvell for releasing the BSD licensed myk(4) driver and
thanks to all users helped fixing bugs.

Tested by: bz, philip, bms,
YAMAMOTO Shigeru < shigeru AT iij DOT ad DOT jp >,
Dmitry Pryanishnikov < dmitry AT atlantis DOT dp DOT ua >,
Jia-Shiun Li < jiashiun AT gmail DOT com >,
David Duchscher < daved AT tamu DOT edu >,
Arno J. Klaassen < arno AT heho DOT snv DOT jussieu DOT fr>,
Nicolae Namolovan < adrenalinup AT gmail DOT com>,
Andre Guibert de Bruet < andy AT siliconlandmark DOT com >
current ML
Tested on: i386, amd64

17 years agoWisdom of Chinggis (Genghis) Khan.
grog [Wed, 13 Dec 2006 00:33:39 +0000 (00:33 +0000)]
Wisdom of Chinggis (Genghis) Khan.

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

17 years agoBe more specific when ELOOP can be returned.
pjd [Tue, 12 Dec 2006 20:06:52 +0000 (20:06 +0000)]
Be more specific when ELOOP can be returned.

17 years agosymlink(2) return EACCES if a component of the name2 path prefix denies
pjd [Tue, 12 Dec 2006 20:05:04 +0000 (20:05 +0000)]
symlink(2) return EACCES if a component of the name2 path prefix denies
write permission.

17 years agoCorrect ENOENT description.
pjd [Tue, 12 Dec 2006 19:57:17 +0000 (19:57 +0000)]
Correct ENOENT description.

17 years agoThe 'name1' argument to symlink(2) is only limited to 1023 characters,
pjd [Tue, 12 Dec 2006 19:48:15 +0000 (19:48 +0000)]
The 'name1' argument to symlink(2) is only limited to 1023 characters,
its components are not checked.

17 years ago- Add constants for HT PCI capability registers including the various
jhb [Tue, 12 Dec 2006 19:33:25 +0000 (19:33 +0000)]
- Add constants for HT PCI capability registers including the various
  subtypes of HT capabilities.
- Add constants for the MSI mapping window HT PCI capability.
- On i386 and amd64, enable the MSI mapping window on any HT bridges we
  encounter and report any non-standard mapping window addresses.

17 years agoGive the WREG() macro the same lifetime as the REG() macro.
jhb [Tue, 12 Dec 2006 19:30:40 +0000 (19:30 +0000)]
Give the WREG() macro the same lifetime as the REG() macro.

17 years agoAdd some bootverbose printf's to detail how many MSI messages are allocated
jhb [Tue, 12 Dec 2006 19:29:01 +0000 (19:29 +0000)]
Add some bootverbose printf's to detail how many MSI messages are allocated
and to which IRQs.

Requested by: scottl

17 years agoGive Host-PCI bridge drivers their own pcib_alloc_msi() and
jhb [Tue, 12 Dec 2006 19:27:01 +0000 (19:27 +0000)]
Give Host-PCI bridge drivers their own pcib_alloc_msi() and
pcib_alloc_msix() methods instead of using the method from the generic
PCI-PCI bridge driver as the PCI-PCI methods will be gaining some PCI-PCI
specific logic soon.

17 years agoSort function prototypes.
jhb [Tue, 12 Dec 2006 19:24:45 +0000 (19:24 +0000)]
Sort function prototypes.

17 years agoReplace a few magic numbers.
jhb [Tue, 12 Dec 2006 19:23:52 +0000 (19:23 +0000)]
Replace a few magic numbers.

17 years agoAdd a function to return the MD interrupt source cookie associated with
jhb [Tue, 12 Dec 2006 19:20:19 +0000 (19:20 +0000)]
Add a function to return the MD interrupt source cookie associated with
an interrupt event.  Use this in the x86 code to fixup the intrcnt names
when an interrupt handler is removed.

17 years agoAdd a comment and fix a whitespace nit.
jhb [Tue, 12 Dec 2006 19:19:22 +0000 (19:19 +0000)]
Add a comment and fix a whitespace nit.

17 years agoIn ip6_sprintf no longer use and return one of eight static buffers
bz [Tue, 12 Dec 2006 17:44:46 +0000 (17:44 +0000)]
In ip6_sprintf no longer use and return one of eight static buffers
for printing/logging ipv6 addresses.
The caller now has to hand in a sufficiently large buffer as first
argument.

This is the "+ one more change" missed in the original commit.

Noticed by: tinderbox
Pointy hat to: me (#1)

17 years ago- Document -u, -i, and the difference between them better.
ru [Tue, 12 Dec 2006 15:26:25 +0000 (15:26 +0000)]
- Document -u, -i, and the difference between them better.
- Remove some historical notes about "future" decisions.

17 years agomkdir(2) creates directory, not file.
pjd [Tue, 12 Dec 2006 15:00:08 +0000 (15:00 +0000)]
mkdir(2) creates directory, not file.

17 years agoCorrectly calculate length of IP header.
glebius [Tue, 12 Dec 2006 12:35:06 +0000 (12:35 +0000)]
Correctly calculate length of IP header.

Submitted by: Eugene Hartmann <eugene tpsb.com.ru>

17 years agoMFp4: 92972, 98913 + one more change
bz [Tue, 12 Dec 2006 12:17:58 +0000 (12:17 +0000)]
MFp4: 92972, 98913 + one more change

In ip6_sprintf no longer use and return one of eight static buffers
for printing/logging ipv6 addresses.
The caller now has to hand in a sufficiently large buffer as first
argument.

17 years agoRemove reference to confusing behaviour just removed from
julian [Tue, 12 Dec 2006 08:13:02 +0000 (08:13 +0000)]
Remove reference to confusing behaviour just removed from
the kse_exit() syscall. Describe the correct behaviour.

17 years agoFix a potential point of confusion. Art Ironport we've seen this end up
julian [Tue, 12 Dec 2006 08:01:55 +0000 (08:01 +0000)]
Fix a potential point of confusion. Art Ironport we've seen this end up
with an infinite loop in and out of the kernel during process shutdown.

17 years agoCorrectly check failed syscall.
davidxu [Tue, 12 Dec 2006 05:26:39 +0000 (05:26 +0000)]
Correctly check failed syscall.

17 years agoFix support for certain 575x/578x chips. This consists of the following:
scottl [Tue, 12 Dec 2006 05:11:12 +0000 (05:11 +0000)]
Fix support for certain 575x/578x chips.  This consists of the following:

- Use the appropriate register writing method when reseting the chip
- Program the descriptor DMA engine correctly.
- More reliably detect certain chips and their features.

Also add some low-level debugging tools to help future work on this driver.

Submitted by: David Christenson (proof of concept changes)
Sponsored by: www.UIA.net

17 years agoworkaround kernel malloc's brittleness
kmacy [Tue, 12 Dec 2006 03:50:06 +0000 (03:50 +0000)]
workaround kernel malloc's brittleness
- don't shuffle phys_avail following kernel to the beginning if the
  range is less than what would remain in a 256MB page (248MB)

17 years agoAdd the '-n' option which is the opposite of '-N', "Do not list tags."
obrien [Tue, 12 Dec 2006 03:20:36 +0000 (03:20 +0000)]
Add the '-n' option which is the opposite of '-N', "Do not list tags."
The '-n' option is needed when one has "log -N" in their ~/.cvsrc, but
wishes to see tags for a particular invocation.

17 years agoMove checking for c_has_waiters into low level _thr_ucond_signal and
davidxu [Tue, 12 Dec 2006 03:08:49 +0000 (03:08 +0000)]
Move checking for c_has_waiters into low level _thr_ucond_signal and
_thr_ucond_broadcast, clear condition variable pointer in cancellation
info after returing from _thr_ucond_wait, since kernel has already
dropped the internal lock, so we don't need to unlock it in cancellation
handler again.

17 years agoIn FreeBSD, we don't need sc_power callback
kevlo [Tue, 12 Dec 2006 03:05:04 +0000 (03:05 +0000)]
In FreeBSD, we don't need sc_power callback

Approved by: cognet and imp

17 years ago- provide a more informative panic if mdesc_update() fails
kmacy [Tue, 12 Dec 2006 02:50:12 +0000 (02:50 +0000)]
- provide a more informative panic if mdesc_update() fails
- handle some cases where the return value of mdesc_update() is not zero
  when it should be

17 years ago- remove vestigial reference to mra[i]
kmacy [Tue, 12 Dec 2006 01:16:17 +0000 (01:16 +0000)]
- remove vestigial reference to mra[i]
- partition phys_avail along 4GB boundaries as possible workaround for hardware
  problems causing watchdog panics

17 years agoAdd MODULE_DEPENDS for cam, pci, mca, eisa and isa where needed.
mjacob [Mon, 11 Dec 2006 23:59:42 +0000 (23:59 +0000)]
Add MODULE_DEPENDS for cam, pci, mca, eisa and isa where needed.

PR: 106543
Reviewed by: Adam Radford
MFC after: 3 days

17 years agoThese days P2P means peer-2-peer (also well known from serveral filesharing
thompsa [Mon, 11 Dec 2006 23:46:40 +0000 (23:46 +0000)]
These days P2P means peer-2-peer (also well known from serveral filesharing
protocols) while PointToPoint has been PtP links. Change the variables
accordingly while the code is still fresh and undocumented.

Requested by: bz

17 years agoNetApp filers return corrupt post op attrs in the wcc on NFS error responses.
mohans [Mon, 11 Dec 2006 19:54:25 +0000 (19:54 +0000)]
NetApp filers return corrupt post op attrs in the wcc on NFS error responses.
This is easy to reproduce for EROFS. I am not sure if the attrs can be corrupt
for other NFS error responses. For now, disabling wcc pre-op attr checks and
post-op attr loads on NFS errors (sysctl'ed).
Reported by: Kris Kennaway

17 years agoFix distorted sound on ASUS P1-AH2 caused by accumulated input / output.
ariff [Mon, 11 Dec 2006 18:45:11 +0000 (18:45 +0000)]
Fix distorted sound on ASUS P1-AH2 caused by accumulated input / output.

Reported/Tested by: Hiroharu Tamaru <tamaru@myn.rcast.u-tokyo.ac.jp>

17 years agoAdd MODULE_DEPENDS for cam, pci, mca, eisa and isa where needed.
mjacob [Mon, 11 Dec 2006 18:28:31 +0000 (18:28 +0000)]
Add MODULE_DEPENDS for cam, pci, mca, eisa and isa where needed.

PR: 106543
MFC after: 3 days

17 years ago- Correct collision counter for BCM5705+. This register is read/clear.
jkim [Mon, 11 Dec 2006 18:00:34 +0000 (18:00 +0000)]
- Correct collision counter for BCM5705+.  This register is read/clear.
- Correct RX packet drop counter for BCM5705+.  This register is read/clear
and it wraps very quickly under heavy packet drops because only the lower
ten bits are valid according to the documentation.  However, it seems few
more bits are actually valid and the rest bits are always zeros[1].
Therefore, we don't mask them off here.  To get accurate packet drop count,
we need to check the register from bge_rxeof().  It is commented out for now,
not to penalize normal operation.  Actual performance impact should be
measured later.
- Correct integer casting from u_long to uint32_t.  Casting is not really
needed for all supported platforms but we better do this correctly[2].

Tested by: bde[1]
Suggested by: bde[2]

17 years agoAdd missing markup bits.
ru [Mon, 11 Dec 2006 11:34:44 +0000 (11:34 +0000)]
Add missing markup bits.

17 years agoo Add support code for newer Marvell PHYs.
yongari [Mon, 11 Dec 2006 11:09:48 +0000 (11:09 +0000)]
o Add support code for newer Marvell PHYs.
o Remove unused static global variable e1000phy_debug.
o Take advantage of mii_phy_dev_probe().
o Use MII_ANEGTICKS/MII_ANEGTICKS_GIGE instead of magic number 5.
o Add IFM_NONE as e1000phy(4) supports it without issues.
o Nuke magic PHY programming sequence in PHY reset and follow correct
  reset sequence. [1]
o Make manual media selection work for all supported media types.
o Don't set MIIF_NOISOLATE so e1000phy(4) can be used in
  configurations with multiple PHYs.
o In 1000baseT, when setting the link manually, one side must be the
  master and the other the slave. If LINK0 is set, program the PHY
  to be a master, otherwise it's a slave.
o When we lost a link, reset mii_ticks immediately so it correctly
  check number of seconds elapsed in autonegotiation phase.
o Announce link loss right after it happens.
o After kicking autonegotiation, report PHY status instead of
  returning immediatly.
o When link state check is in progress, check auto negotiation
  completion bit only when auto negotiation is enbaled.
o When PHY is resolved to a master, show it with IFM_FLAG2.

Special thanks to marius who fixed several nits in original patch.
In half-duplex mode, nfe(4) fails to send packets. I think it's a bug
in nfe(4) as the same PHY works without problems on msk(4).

Obtained from: em(4) [1]
Reviewed by: marius
Tested by: bz

17 years agoFix a typo ceri found while reviewing and I forgot to commit
erwin [Mon, 11 Dec 2006 11:01:13 +0000 (11:01 +0000)]
Fix a typo ceri found while reviewing and I forgot to commit
in the previous version.

Submitted by: ceri, matteo
Reviewed by: cperciva

Also note that the last revision was:
Reviewed by: ceri, simon

17 years agoAdd additional Marvell PHY registers.
yongari [Mon, 11 Dec 2006 10:43:32 +0000 (10:43 +0000)]
Add additional Marvell PHY registers.

17 years agoAdd a bunch of new Marvell PHY ids.
yongari [Mon, 11 Dec 2006 10:42:04 +0000 (10:42 +0000)]
Add a bunch of new Marvell PHY ids.

17 years agostyle(9)
yongari [Mon, 11 Dec 2006 10:22:24 +0000 (10:22 +0000)]
style(9)

17 years agodefer all processing to a full fledged thread.
alfred [Mon, 11 Dec 2006 09:57:41 +0000 (09:57 +0000)]
defer all processing to a full fledged thread.

once usb is SMP safe, this should be the first SMPsafe
usb ethernet driver.

17 years agoDocument the ELF types supported by libelf.
jkoshy [Mon, 11 Dec 2006 09:03:27 +0000 (09:03 +0000)]
Document the ELF types supported by libelf.

Document the relationship between section types (SHT_*) and the
types known to the libelf.

17 years agoUpdate the list of ports categories.
erwin [Mon, 11 Dec 2006 08:54:17 +0000 (08:54 +0000)]
Update the list of ports categories.

This is an updated version of:
PR:             85367
Submitted by:   linimon

Reviewed by:    simon
MFC after:      3 days

17 years agomake size of pad non-zero so that trap-tracing code doesn't overwrite the
kmacy [Mon, 11 Dec 2006 04:50:25 +0000 (04:50 +0000)]
make size of pad non-zero so that trap-tracing code doesn't overwrite the
base of our stack

17 years agoTeach the brgphy about some new Broadcom IDs.
scottl [Mon, 11 Dec 2006 01:29:40 +0000 (01:29 +0000)]
Teach the brgphy about some new Broadcom IDs.

Submitted by: David Christenson
Sponsored by: www.UIA.net

17 years agoreplace if_watchdog w/ a private callout
sam [Mon, 11 Dec 2006 00:41:22 +0000 (00:41 +0000)]
replace if_watchdog w/ a private callout

MFC after: 1 month

17 years agodrop softc lock when passing rx frames up the stack
sam [Mon, 11 Dec 2006 00:37:12 +0000 (00:37 +0000)]
drop softc lock when passing rx frames up the stack

MFC after: 1 month

17 years agosplit wi_start int locked+unlocked variants and use the unlocked
sam [Mon, 11 Dec 2006 00:35:51 +0000 (00:35 +0000)]
split wi_start int locked+unlocked variants and use the unlocked
one from the isr to eliminate a recursive lock

MFC after: 1 month

17 years agonuke compatibility cruft
sam [Mon, 11 Dec 2006 00:32:15 +0000 (00:32 +0000)]
nuke compatibility cruft

MFC after: 1 month

17 years agoKTR entry contained invalid context reference - ifdef out
kmacy [Sun, 10 Dec 2006 18:09:44 +0000 (18:09 +0000)]
KTR entry contained invalid context reference - ifdef out

17 years agoAdd a new turkish mirror
erwin [Sun, 10 Dec 2006 16:33:38 +0000 (16:33 +0000)]
Add a new turkish mirror

Submitted by:   Evren Yurtesen <yurtesen@ispro.net>
Reviewed by:    simon
MFC after:      3 days

17 years agoBack out revision 1.264.
bms [Sun, 10 Dec 2006 13:44:00 +0000 (13:44 +0000)]
Back out revision 1.264.

Fixing the IP accounting issue, if we plan to do so, needs to be better
thought out; the 'fix' introduces a hash lookup and a possible kernel panic.

Reported by: Mark Tinguely

17 years agoWhen directory is given as an argument for unlink(2), EPERM is returned
pjd [Sun, 10 Dec 2006 13:04:36 +0000 (13:04 +0000)]
When directory is given as an argument for unlink(2), EPERM is returned
not matter if this is regular directory or a mount point.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r165078,
dougb [Sun, 10 Dec 2006 07:12:50 +0000 (07:12 +0000)]
This commit was generated by cvs2svn to compensate for changes in r165078,
which included commits to RCS files with non-trunk default branches.

17 years agoChanges to generated files related to the 9.3.3 import.
dougb [Sun, 10 Dec 2006 07:11:04 +0000 (07:11 +0000)]
Changes to generated files related to the 9.3.3 import.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r165071,
dougb [Sun, 10 Dec 2006 07:09:56 +0000 (07:09 +0000)]
This commit was generated by cvs2svn to compensate for changes in r165071,
which included commits to RCS files with non-trunk default branches.

17 years agoVendor import of BIND 9.3.3
dougb [Sun, 10 Dec 2006 07:09:56 +0000 (07:09 +0000)]
Vendor import of BIND 9.3.3

17 years agoStyle: Shorten a couple of lines with u_int and u_long.
imp [Sun, 10 Dec 2006 06:36:41 +0000 (06:36 +0000)]
Style: Shorten a couple of lines with u_int and u_long.

17 years agoFix AD1986A (and possibily other codecs too) microphone issues.
ariff [Sun, 10 Dec 2006 06:13:41 +0000 (06:13 +0000)]
Fix AD1986A (and possibily other codecs too) microphone issues.

Tested by: xride (using skype, xanalyser, etc. ; play/record)

17 years agoremove more uses of trap_conversion to get more meaningful trap messages
kmacy [Sun, 10 Dec 2006 06:00:09 +0000 (06:00 +0000)]
remove more uses of trap_conversion to get more meaningful trap messages
add a printf for when we fault on the direct area (should never happen)

17 years agoRemember to include isp_library.o when building an isp module.
mjacob [Sun, 10 Dec 2006 05:51:27 +0000 (05:51 +0000)]
Remember to include isp_library.o when building an isp module.

17 years agoMake cam_xpt's pronouncements match camcontrol
mjacob [Sun, 10 Dec 2006 05:37:48 +0000 (05:37 +0000)]
Make cam_xpt's pronouncements match camcontrol
(Tagged -> Command) Queueing.

17 years agoMake MAXPHYS and DFLTPHYS options (finally).
mjacob [Sun, 10 Dec 2006 04:23:23 +0000 (04:23 +0000)]
Make MAXPHYS and DFLTPHYS options (finally).

17 years agoDo explicit bounds checking as a function of the actual size of the
kmacy [Sun, 10 Dec 2006 04:18:03 +0000 (04:18 +0000)]
Do explicit bounds checking as a function of the actual size of the
reloc_target_bitmask array as opposed to the (known) index of the last value.
This change fixes CID 691.

17 years agobetter handle the case of hw.physmemstart being hw.physmem not being set,
kmacy [Sun, 10 Dec 2006 04:14:29 +0000 (04:14 +0000)]
better handle the case of hw.physmemstart being hw.physmem not being set,
previously we were acting as if physmem was being set when it was not

17 years agoThe MODULE_NAME should "ispfw" unless overriden, not "isp".
mjacob [Sun, 10 Dec 2006 03:42:57 +0000 (03:42 +0000)]
The MODULE_NAME should "ispfw" unless overriden, not "isp".

17 years agoRemove dependency on ispfw and firmware as modules.
mjacob [Sun, 10 Dec 2006 03:41:48 +0000 (03:41 +0000)]
Remove dependency on ispfw and firmware as modules.
Either they're there early and the ispfw sets have
registered themselves, or they're not.

The module dependency stuff isn't quite what we want
anyway. If the user doesn't want the load placed on
system memory by loading the firmware, they don't
specify it to be loaded (either by being linked in
or via being a module to be loaded and then hooked
in with firmware(9)). It doesn't then make sense to
then override what they want by pulling it in anyway.

This might be able to work if we were able to pull in
just exactly what we needed for the card we have- but
that's an optimization left for the future.

17 years agoAdd hw.physmemstart loader variable to enable the user to specify the address
kmacy [Sun, 10 Dec 2006 01:52:46 +0000 (01:52 +0000)]
Add hw.physmemstart loader variable to enable the user to specify the address
at which the kernel should start allocating physical memory. The primary
purpose of this is to test 64-bit cleanness of the data path by setting
hw.physmemstart=4G so that all physical allocations are above 4GB. AMD64
and i386/PAE could also benefit from having this option.

17 years agoMake mpt_pci depend on pci and mpt_cam depend on CAM.
mjacob [Sun, 10 Dec 2006 01:13:56 +0000 (01:13 +0000)]
Make mpt_pci depend on pci and mpt_cam depend on CAM.

PR: 106536
Suggested by: Norikatsu Shigemura
MFC after: 3 days

17 years agoAs Bernd Walter points out, the rlphy is used for more things than
imp [Sun, 10 Dec 2006 01:10:08 +0000 (01:10 +0000)]
As Bernd Walter points out, the rlphy is used for more things than
just the intenral phy on parts supported by the rl and re drivers, the
RTL8201BL for example.  He also sent me a nice picture of hundreds of
these chips in a tray to boulder his claim.  :-) Therefore remove a
comment that suggested that they were...

17 years agoIf the named file has its immutable or append-only flag set, utimes(2)
pjd [Sat, 9 Dec 2006 23:13:05 +0000 (23:13 +0000)]
If the named file has its immutable or append-only flag set, utimes(2)
returns EPERM.