]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
13 years agoAdd support to Alcatel/TCTMobile X080S USB 3G modem. The device needs
glebius [Wed, 6 Oct 2010 14:29:00 +0000 (14:29 +0000)]
Add support to Alcatel/TCTMobile X080S USB 3G modem. The device needs
special eject command to reappear as modem. It also requires DIR_IN flag
in the command message, so we supply some dummy data along with the command.

Feedback from X080S owners appreciated. I have not a pure Alcatel/TCTMobile
device, but another one under "Svyaznoy" (Связной) brand, and I didn't yet
managed to get it working. It is successfully recognized, it responds to
AT commands, but it shuts up right after successfull CONNECT response.

Reviewed by: hps

13 years agoFix regex for some socket- and acl-related syscall values.
netchild [Wed, 6 Oct 2010 12:41:42 +0000 (12:41 +0000)]
Fix regex for some socket- and acl-related syscall values.

Submitted by: Mikhail <hidden.node@gmail.com>
PR: 149295
MFC-after: 2 weeks

13 years agoMissed space.
kib [Wed, 6 Oct 2010 09:24:33 +0000 (09:24 +0000)]
Missed space.

Submitted by: brueffer
MFC after: 1 week

13 years agoAdd cross-references to lrand48(3) and arc4random(3) from rand(3)
kib [Wed, 6 Oct 2010 09:05:47 +0000 (09:05 +0000)]
Add cross-references to lrand48(3) and arc4random(3) from rand(3)
and random(3).

Submitted by: Valentin Nechayev <netch netch kiev ua>
MFC after: 1 week

13 years agoXLR/XLS network driver (nlge) updates:
jchandra [Wed, 6 Oct 2010 08:15:28 +0000 (08:15 +0000)]
XLR/XLS network driver (nlge) updates:

- nlge_ioctl handles IFF_UP and IFF_PROMISC flags
- Translate table code, to enable flow based CPU assignment added
  disabled by default (can be enabled by a tunable).
- Changed signature of nlge_port_disable to make it consistent with nlge_port_enable
- Removed TXCSUM and VLAN_HW_TAGGING from i/f capabilities.

Submitted by: Sriram Gorti (srgorti at netlogicmicro dot com)

13 years agoFix n64 compile.
jchandra [Wed, 6 Oct 2010 08:09:39 +0000 (08:09 +0000)]
Fix n64 compile.

13 years agoAdd some missing files to optionally delete.
netchild [Wed, 6 Oct 2010 07:49:28 +0000 (07:49 +0000)]
Add some missing files to optionally delete.

Submitted by: Paul B Mahol <onemda@gmail.com>

13 years agoFix a comparision of an uninitialised pointer.
netchild [Wed, 6 Oct 2010 07:34:41 +0000 (07:34 +0000)]
Fix a comparision of an uninitialised pointer.

Submitted by: arundel
Found by: clang analysis (automatic service by uqs@)
Reviewed by: rdivacky

13 years agoIf LANG/LC_CTYPE/LC_ALL is set and the localized man page contains a page
gordon [Wed, 6 Oct 2010 07:22:56 +0000 (07:22 +0000)]
If LANG/LC_CTYPE/LC_ALL is set and the localized man page contains a page
also in the base set, the man utility when invoked with -a would display
pages for each locale in the same tree:

$ LANG=en_GB.ISO8859-15 man -wa man
/usr/share/man/en.ISO8859-15/man1/man.1.gz
/usr/share/man/man1/man.1.gz
/usr/share/man/en.ISO8859-15/man7/man.7.gz
/usr/share/man/man7/man.7.gz

Use continue to break out of the loop for the current locale. This results
in behavior more closely matching the old GNU man implementation:

$ LANG=en_GB.ISO8859-15 man -wa man
/usr/share/man/en.ISO8859-15/man1/man.1.gz
/usr/share/man/en.ISO8859-15/man7/man.7.gz

This will still search for a copy of the file in other manual path
locations. If there was a /usr/local/man/man1/man.1.gz file, it would still
be displayed. This is also consistent with the GNU man implementation.

Submitted by: arundel
Approved by: wes (mentor implicit)

13 years agoRecognize both ! and ^ as markers for negated character classes.
kientzle [Wed, 6 Oct 2010 04:30:40 +0000 (04:30 +0000)]
Recognize both ! and ^ as markers for negated character classes.

Submitted by: Mykola Dzham

13 years agoFix bge(4) build breakage when BGE_REGISTER_DEBUG is defined.
yongari [Wed, 6 Oct 2010 01:23:40 +0000 (01:23 +0000)]
Fix bge(4) build breakage when BGE_REGISTER_DEBUG is defined.

13 years agoRearrange code a bit to correctly set PHY flags. This change make
yongari [Tue, 5 Oct 2010 23:24:58 +0000 (23:24 +0000)]
Rearrange code a bit to correctly set PHY flags. This change make
it easy to add more newer ASICs.

Obtained from: OpenBSD

13 years agoSeparate common flags into controller specific and PHY related
yongari [Tue, 5 Oct 2010 23:03:48 +0000 (23:03 +0000)]
Separate common flags into controller specific and PHY related
flags. There should be no functional changes. This change will make
it easy to add more quirk/flags in future.

Reviewed by: davidch

13 years ago- Add WITH_GPIO entry to src.conf(5) man page
gonzo [Tue, 5 Oct 2010 22:26:01 +0000 (22:26 +0000)]
- Add WITH_GPIO entry to src.conf(5) man page

13 years ago- Introduce WITH_GPIO knob and disable building gpioctl
gonzo [Tue, 5 Oct 2010 21:19:20 +0000 (21:19 +0000)]
- Introduce WITH_GPIO knob and disable building gpioctl
    by default. Most systems do not need it.

13 years agoUse the printf-like capability from kproc_create().
thompsa [Tue, 5 Oct 2010 20:56:08 +0000 (20:56 +0000)]
Use the printf-like capability from kproc_create().

Submitted by: Paul B Mahol

13 years agoSEE ALSO should sorted by manpage section first.
brueffer [Tue, 5 Oct 2010 20:39:36 +0000 (20:39 +0000)]
SEE ALSO should sorted by manpage section first.

13 years agoRevert the unrelated patch crept in the previous commit.
jkim [Tue, 5 Oct 2010 18:47:51 +0000 (18:47 +0000)]
Revert the unrelated patch crept in the previous commit.

13 years agoPrefer pmap_unmapbios() over pmap_unmapdev(). The binary does not change
jkim [Tue, 5 Oct 2010 18:38:23 +0000 (18:38 +0000)]
Prefer pmap_unmapbios() over pmap_unmapdev().  The binary does not change
after this because pmap_unmapbios() is a macro for pmap_unmapdev() on amd64.

13 years agoHandle vector assist traps without a kernel panic, by setting denormalized
nwhitehorn [Tue, 5 Oct 2010 18:08:07 +0000 (18:08 +0000)]
Handle vector assist traps without a kernel panic, by setting denormalized
values to zero. A correct solution would involve emulating vector
operations on denormalized values, but this has little effect on accuracy
and is much less complicated for now.

MFC after: 2 weeks

13 years agoInitialize KPTmap in locore so that vm86.c can call vtophys() (or really
alc [Tue, 5 Oct 2010 17:06:51 +0000 (17:06 +0000)]
Initialize KPTmap in locore so that vm86.c can call vtophys() (or really
pmap_kextract()) before pmap_bootstrap() is called.

Document the set of pmap functions that may be called before
pmap_bootstrap() is called.

Tested by: bde@
Reviewed by: kib@
Discussed with: jhb@
MFC after: 6 weeks

13 years agoAdd reference to intro(1) in SEE ALSO section.
gordon [Tue, 5 Oct 2010 16:14:08 +0000 (16:14 +0000)]
Add reference to intro(1) in SEE ALSO section.

PR: docs/151216
Approved by: wes (mentor, implicit)

13 years agoWhen no protocol entry is found, getproto*_r(3) should
ume [Tue, 5 Oct 2010 15:40:59 +0000 (15:40 +0000)]
When no protocol entry is found, getproto*_r(3) should
return zero.

Spotted by: Kostik Belousov <kostikbel__at__gmail.com>
MFC after: 2 weeks

13 years agoDisplay PCID capability of CPU and add CPUID define for it.
kib [Tue, 5 Oct 2010 15:31:56 +0000 (15:31 +0000)]
Display PCID capability of CPU and add CPUID define for it.

MFC after: 1 week

13 years agoHandle null return from XML_ParserCreate and fix a few memory leaks on
emaste [Tue, 5 Oct 2010 15:27:44 +0000 (15:27 +0000)]
Handle null return from XML_ParserCreate and fix a few memory leaks on
error conditions.

Submitted by: Mark Johnston <mjohnston at sandvine dot com>
MFC after: 2 weeks

13 years agoNetwork driver support for more XLR/XLS chip versions.
jchandra [Tue, 5 Oct 2010 07:41:59 +0000 (07:41 +0000)]
Network driver support for more XLR/XLS chip versions.

Obtained from: Sriram Gorti (srgorti at netlogicmicro dot com)

13 years agoClear any pending interrupts after disabling FMN interrupts. Add debugging
jchandra [Tue, 5 Oct 2010 06:44:47 +0000 (06:44 +0000)]
Clear any pending interrupts after disabling FMN interrupts. Add debugging
sysctl and counters for message ring threads (intial version). Update
watermark values, and and decrease the maximum threads to 3 (this will leave
a few CPUs for other processes)

Minor comment fix in nlge.

13 years agoRevise markup, mainly.
ru [Tue, 5 Oct 2010 06:31:11 +0000 (06:31 +0000)]
Revise markup, mainly.

13 years agoFix asm for XKPHYS read for o32, constriants needed for input registers
jchandra [Tue, 5 Oct 2010 05:49:38 +0000 (05:49 +0000)]
Fix asm for XKPHYS read for o32, constriants needed for input registers
that are clobbered.  Fixes a crash when compiled without DEBUG

13 years agoShow hostcache statistics.
ru [Tue, 5 Oct 2010 05:15:27 +0000 (05:15 +0000)]
Show hostcache statistics.

Submitted by: Maxim Dounin

13 years agoPrint out correct USB connection speed for USB 3.0 mass storage devices.
hselasky [Mon, 4 Oct 2010 23:30:44 +0000 (23:30 +0000)]
Print out correct USB connection speed for USB 3.0 mass storage devices.

Approved by:    thompsa (mentor)

13 years agoRX buffer allocation failure is not an input error. Controller
yongari [Mon, 4 Oct 2010 23:25:38 +0000 (23:25 +0000)]
RX buffer allocation failure is not an input error. Controller
successfully received a frame but we failed to pass it to upper
stack due to lack of resources. So update if_iqdrops counter
instead of updating if_ierrors counter.

13 years agoLink the XHCI (USB 3.0 hardware driver) into the default kernel build.
hselasky [Mon, 4 Oct 2010 23:22:03 +0000 (23:22 +0000)]
Link the XHCI (USB 3.0 hardware driver) into the default kernel build.

Approved by:    thompsa (mentor)

13 years agoDon't count input error twice. uether_rxbuf() already updated that
yongari [Mon, 4 Oct 2010 23:19:31 +0000 (23:19 +0000)]
Don't count input error twice. uether_rxbuf() already updated that
counter.

13 years agoThis commit adds full support for USB 3.0 devices in host and device
hselasky [Mon, 4 Oct 2010 23:18:05 +0000 (23:18 +0000)]
This commit adds full support for USB 3.0 devices in host and device
mode in the USB core.  The patch mostly consists of updating the USB
HUB code to support USB 3.0 HUBs. This patch also add some more USB
controller methods to support more active-alike USB controllers like
the XHCI which needs to be informed about various device state events.

USB 3.0 HUBs are not tested yet, due to lack of hardware, but are
believed to work.

After this update the initial device descriptor is only read twice
when we know that the bMaxPacketSize is too small for a single packet
transfer of this descriptor.

Approved by:    thompsa (mentor)

13 years agoAdd missing #if's
hselasky [Mon, 4 Oct 2010 22:50:08 +0000 (22:50 +0000)]
Add missing #if's

Approved by:    thompsa (mentor)

13 years agoAdd more strict USB string filtering.
hselasky [Mon, 4 Oct 2010 22:45:17 +0000 (22:45 +0000)]
Add more strict USB string filtering.

Approved by:    thompsa (mentor)

13 years agoSerialise USB re-enumeration with the USB explore thread.
hselasky [Mon, 4 Oct 2010 22:21:30 +0000 (22:21 +0000)]
Serialise USB re-enumeration with the USB explore thread.
This patch can solve problems when multiple USB devices are
re-enumerated at the same time on the same bus.

Approved by:    thompsa (mentor)

13 years agoCorrect IOCTL return code.
hselasky [Mon, 4 Oct 2010 22:04:22 +0000 (22:04 +0000)]
Correct IOCTL return code.

Approved by:    thompsa (mentor)

13 years agoDecrease report interval to 5 seconds, as this also means we will check for
pjd [Mon, 4 Oct 2010 21:44:26 +0000 (21:44 +0000)]
Decrease report interval to 5 seconds, as this also means we will check for
signals every 5 seconds and not every 10 seconds as before.

MFC after: 3 days

13 years agohook_check() is now only used to report about long-running hooks, so the
pjd [Mon, 4 Oct 2010 21:43:06 +0000 (21:43 +0000)]
hook_check() is now only used to report about long-running hooks, so the
argument is redundant, remove it.

MFC after: 3 days

13 years agoWe can't mask ignored signal, so install dummy signal hander for SIGCHLD before
pjd [Mon, 4 Oct 2010 21:41:18 +0000 (21:41 +0000)]
We can't mask ignored signal, so install dummy signal hander for SIGCHLD before
masking it.

This fixes bogus reports about hooks running for too long and other problems
related to garbage-collecting child processes.

Reported by: Mikolaj Golub <to.my.trociny@gmail.com>
MFC after: 3 days

13 years agoAdd missing USB 3.0 definitions. Correct some wrong ones.
hselasky [Mon, 4 Oct 2010 21:38:10 +0000 (21:38 +0000)]
Add missing USB 3.0 definitions. Correct some wrong ones.

Approved by:    thompsa (mentor)

13 years agoAdd missing DRIVER_MODULE() entry for the musbotg driver.
hselasky [Mon, 4 Oct 2010 21:30:15 +0000 (21:30 +0000)]
Add missing DRIVER_MODULE() entry for the musbotg driver.
Add some more comments.

Approved by:    thompsa (mentor)

13 years agoThe root mount hold reference was not released on USB controller
hselasky [Mon, 4 Oct 2010 21:24:10 +0000 (21:24 +0000)]
The root mount hold reference was not released on USB controller
attach failures during boot. Fix this.

Approved by:    thompsa (mentor)

13 years agoMake upper stack know driver's output status. This change increased
yongari [Mon, 4 Oct 2010 21:01:27 +0000 (21:01 +0000)]
Make upper stack know driver's output status. This change increased
TX performance from 221kpps to 231kpps.

13 years agoMove updating TX packet counter to the inside of send loop. axe(4)
yongari [Mon, 4 Oct 2010 20:49:38 +0000 (20:49 +0000)]
Move updating TX packet counter to the inside of send loop. axe(4)
controllers combine multiple TX requests into single one if there
is room in TX buffer of controller. Updating TX packet counter at
the end of TX completion resulted in incorrect TX packet counter as
axe(4) thought it sent 1 packet. There is no easy way to know how
many combined TX were completed in the callback.
Because this change updates TX packet counter before actual
transmission, it may not be ideal one. But I believe it's better
than showing fake 8kpps under high TX load. With this change, TX
shows 221kpps on Linksus USB200M.

13 years agoChange libvgl's set4pixels() and set2lines() functions from plain
dim [Mon, 4 Oct 2010 18:16:38 +0000 (18:16 +0000)]
Change libvgl's set4pixels() and set2lines() functions from plain
'inline' to 'static inline'.  Otherwise, a C99 compiler (such as clang)
will output an undefined symbol for those functions in the resulting
object file.  (Even gcc will do this, when you use "-std=c99".)

This should fix the "undefined reference to `set4pixels'" errors that
some people were seeing during ports building, when their world was
compiled with clang.

Approved by: rpaulo (mentor)

13 years agoEnable fix for read DMA FIFO overruns on controllers that have this
yongari [Mon, 4 Oct 2010 18:09:01 +0000 (18:09 +0000)]
Enable fix for read DMA FIFO overruns on controllers that have this
fix. Note, we still need workaround for controllers that lacks this
fix and it needs more work in RX BD updating.

Submitted by: davidch

13 years agoConsistently use ifHCOutOctets/ifHCInOctets instead of Octets as
yongari [Mon, 4 Oct 2010 18:01:23 +0000 (18:01 +0000)]
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as
these names are used in data sheet. Also use UnicastPkts,
MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts
and BcastPkts to clarify its meaning.

Suggested by: bde

13 years agoFix two aliases that had the same name but were pointing to different
gnn [Mon, 4 Oct 2010 17:22:18 +0000 (17:22 +0000)]
Fix two aliases that had the same name but were pointing to different
events.  These are now disamiguated.

MFC after: 1 week

13 years agoIf vm_map_find() is asked to allocate a superpage-aligned region of virtual
alc [Mon, 4 Oct 2010 16:49:40 +0000 (16:49 +0000)]
If vm_map_find() is asked to allocate a superpage-aligned region of virtual
addresses that is greater than a superpage in size but not a multiple of
the superpage size, then vm_map_find() is not always expanding the kernel
pmap to support the last few small pages being allocated.  These failures
are not commonplace, so this was first noticed by someone porting FreeBSD
to a new architecture.  Previously, we grew the kernel page table in
vm_map_findspace() when we found the first available virtual address.
This works most of the time because we always grow the kernel pmap or page
table by an amount that is a multiple of the superpage size.  Now, instead,
we defer the call to pmap_growkernel() until we are committed to a range
of virtual addresses in vm_map_insert().  In general, there is another
reason to prefer calling pmap_growkernel() in vm_map_insert().  It makes
it possible for someone to do the equivalent of an mmap(MAP_FIXED) on the
kernel map.

Reported by: Svatopluk Kraus
Reviewed by: kib@
MFC after: 3 weeks

13 years agoFollow exactly the steps in architecture manual for correctly invalidating
nwhitehorn [Mon, 4 Oct 2010 16:07:48 +0000 (16:07 +0000)]
Follow exactly the steps in architecture manual for correctly invalidating
TLB entries instead of trying to cut corners.

13 years agoFix two subtle problems in PPC32 RTLD. The first is a concurrency issue
nwhitehorn [Mon, 4 Oct 2010 16:02:11 +0000 (16:02 +0000)]
Fix two subtle problems in PPC32 RTLD. The first is a concurrency issue
where long PLT calls in multi-threaded environments could end up with
incorrect jmptab values. The second is that, after the addition of extended
PLT support, I forgot to update the PLT icache synchronization code to cover
the extended PLT instead of just the basic PLT.

MFC after: 10 days

13 years agoAdd updating entry for manpath.config deprecation.
gordon [Mon, 4 Oct 2010 15:39:53 +0000 (15:39 +0000)]
Add updating entry for manpath.config deprecation.

Approved by: wes (mentor)

13 years agoClear errno for each method dispatch.
ume [Mon, 4 Oct 2010 15:28:01 +0000 (15:28 +0000)]
Clear errno for each method dispatch.

Spotted by: Kostik Belousov <kostikbel__at__gmail.com>
MFC after: 2 weeks

13 years agoFix punctuation and grammar, mostly by ending sentences with a period.
gnn [Mon, 4 Oct 2010 14:32:14 +0000 (14:32 +0000)]
Fix punctuation and grammar, mostly by ending sentences with a period.

MFC after: 1 day

13 years agoCertain static code analysis tools (FlexeLint being one) are very
phk [Mon, 4 Oct 2010 10:48:47 +0000 (10:48 +0000)]
Certain static code analysis tools (FlexeLint being one) are very
suspicious about 'l' and '1' being confused in numeric constants.
The fear being that some old fart programmer might still think that
he is using a Remmington Noiseless as input terminal device.

An easy way to placate this fear is to use capital 'L' or to put
the 'u' in unsigned constants in front of the 'l'.

13 years agoClarify the combination effect of -P and -f to make it clear.
delphij [Mon, 4 Oct 2010 06:17:45 +0000 (06:17 +0000)]
Clarify the combination effect of -P and -f to make it clear.

Submitted by: arundel
MFC after: 2 weeks

13 years agoBump __FreeBSD_version for switch from GNU man to BSDL man.
gordon [Mon, 4 Oct 2010 01:08:31 +0000 (01:08 +0000)]
Bump  __FreeBSD_version for switch from GNU man to BSDL man.

Approved by: wes (mentor)

13 years agoNo longer install /etc/manpath.config since the BSDL man utilities
gordon [Mon, 4 Oct 2010 01:07:04 +0000 (01:07 +0000)]
No longer install /etc/manpath.config since the BSDL man utilities
don't use it.

Approved by: wes (mentor)

13 years agoRetire GNU man in favor of the newly written BSDL version.
gordon [Sun, 3 Oct 2010 22:24:14 +0000 (22:24 +0000)]
Retire GNU man in favor of the newly written BSDL version.

Approved by: wes (mentor)

13 years agoReconnect uathload to the build.
antoine [Sun, 3 Oct 2010 20:09:19 +0000 (20:09 +0000)]
Reconnect uathload to the build.

13 years agoUnbreak build.
antoine [Sun, 3 Oct 2010 20:04:11 +0000 (20:04 +0000)]
Unbreak build.

13 years agoAdd more obsolete files.
antoine [Sun, 3 Oct 2010 18:24:31 +0000 (18:24 +0000)]
Add more obsolete files.

13 years agoExtend sysinstall to handle powerpc64 by teaching it that powerpc64 has
nwhitehorn [Sun, 3 Oct 2010 17:50:43 +0000 (17:50 +0000)]
Extend sysinstall to handle powerpc64 by teaching it that powerpc64 has
a lib32 distribution, and that the GENERIC kernel is named GENERIC64.
More modifications will be required later for installations from ftp due
to the shared platform name with 32-bit powerpc, but this is enough for
snapshot CDs to work.

Reviewed by: brucec

13 years agoRemove an header that apart from the license is empty.
marius [Sun, 3 Oct 2010 17:05:25 +0000 (17:05 +0000)]
Remove an header that apart from the license is empty.

13 years agoConsistently always explicitly set IFM_HDX for half-duplex.
marius [Sun, 3 Oct 2010 17:00:57 +0000 (17:00 +0000)]
Consistently always explicitly set IFM_HDX for half-duplex.

Obtained from: OpenBSD (mostly)

13 years agoAdd a memory-range interface to /dev/mem on PowerPC using PAT attributes.
nwhitehorn [Sun, 3 Oct 2010 16:02:53 +0000 (16:02 +0000)]
Add a memory-range interface to /dev/mem on PowerPC using PAT attributes.
Unlike actual MTRR, this only controls the mapping attributes for
subsequent mmap() of /dev/mem. Nonetheless, the support is sufficiently
MTRR-like that Xorg can use it, which translates into an enormous increase
in graphics performance on PowerPC.

MFC after: 2 weeks

13 years agoThe makectx() function, used by kdb_trap() to reconstruct pcb from
kib [Sun, 3 Oct 2010 13:52:17 +0000 (13:52 +0000)]
The makectx() function, used by kdb_trap() to reconstruct pcb from
trap frame when trap initiated kdb entry, incorrectly calculated the
value of %rsp for trapped thread.

According to Intel(R) 64 and IA-32 Architectures Software Developer's Manual
Volume 3A: System Programming Guide, Part 1, rev. 035, 6.14.2 64-Bit Mode
Stack Frame, "64-bit mode ... pushes SS:RSP unconditionally, rather than
only on a CPL change."
Even assuming the conditional push of the %ss:%rsp, the calculation
was still wrong because sizeof(tf_ss) + sizeof(tf_rsp) == 16 on amd64.

Always use the tf_rsp from trap frame. The change supposedly fixes
stepping when using kgdb backend for kdb.

Submitted by: Zhouyi Zhou <zhouzhouyi gmail com>
PR: amd64/151167
Reviewed by: avg
MFC after: 1 week

13 years agoReplace an obsolete flag -L in an mkisofs(1) command line with
hrs [Sun, 3 Oct 2010 13:13:10 +0000 (13:13 +0000)]
Replace an obsolete flag -L in an mkisofs(1) command line with
-allow-leading-dots to fix "make release" for FreeBSD/powerpc.

Reviewed by: marcel

13 years agoCommit initial version of new XHCI driver which was written from
hselasky [Sun, 3 Oct 2010 08:12:17 +0000 (08:12 +0000)]
Commit initial version of new XHCI driver which was written from
scratch. This driver adds support for USB3.0 devices. The XHCI
interface is also backwards compatible to USB2.0 and USB1.0 and will
evntually replace the OHCI/UHCI and EHCI drivers.

There will be follow-up commits during the coming week to link the
driver into the default kernel build and add missing USB3.0
functionality in the USB core. Currently only the driver files are
committed.

Approved by: thompsa (mentor)

13 years agoUpdate message ring handling code for XLR/XLS
jchandra [Sun, 3 Oct 2010 04:33:58 +0000 (04:33 +0000)]
Update message ring handling code for XLR/XLS

- Wakeup multiple threads per core using message ring watermark interrupts.
- Update message ring handler registration, use the real device station id
  for registering interrupts.
- rge/nlge: update for the new message ring registration code.
- rge/nlge: use 2 message ring stations for incoming packets, this will
  allow more messages to be queued.
- nlge: comment fixes, remove unused variable
- style and whitespace fixes

13 years agoAdjust the all target message (but maybe all: sysent is better?
imp [Sat, 2 Oct 2010 22:12:41 +0000 (22:12 +0000)]
Adjust the all target message (but maybe all: sysent is better?

13 years agoTurns out this file was how we make sysent stuff, so add that part only back...
imp [Sat, 2 Oct 2010 21:35:33 +0000 (21:35 +0000)]
Turns out this file was how we make sysent stuff, so add that part only back...

13 years agoSplit the root mount logic from the (generic) mount code and move
marcel [Sat, 2 Oct 2010 19:44:13 +0000 (19:44 +0000)]
Split the root mount logic from the (generic) mount code and move
it (the root mount code) into a new file called vfs_mountroot.c

The split is almost trivial, as the code is almost perfectly
non-intertwined. The only adjustment needed was to move the UMA
zone allocation out of vfs_mountroot() [in vfs_mountroot.c] and
into vfs_mount.c, where it had to be done as a SYSINIT [see
vfs_mount_init()].

There are no functional changes with this commit.

13 years ago- In the spirit of previous simplifications factor out the checks for a
marius [Sat, 2 Oct 2010 18:53:12 +0000 (18:53 +0000)]
- In the spirit of previous simplifications factor out the checks for a
  different PHY instance being selected and isolation out into the wrappers
  around the service methods rather than duplicating them over and over
  again (besides, a PHY driver shouldn't need to care about which instance
  it actually is).
- Centralize the check for the need to isolate a non-zero PHY instance not
  supporting isolation in mii_mediachg() and just ignore it rather than
  panicing, which should sufficient given that a) things are likely to
  just work anyway if one doesn't plug in more than one port at a time and
  b) refusing to attach in this case just leaves us in a unknown but most
  likely also not exactly correct configuration (besides several drivers
  setting MIIF_NOISOLATE didn't care about these anyway, probably due to
  setting this flag for no real reason).
- Minor fixes like removing unnecessary setting of sc->mii_anegticks,
  using sc->mii_anegticks instead of hardcoded values etc.

13 years agoM_USE_RESERVE has been deprecated for a decade. Eliminate any uses that
alc [Sat, 2 Oct 2010 17:58:57 +0000 (17:58 +0000)]
M_USE_RESERVE has been deprecated for a decade.  Eliminate any uses that
have no run-time effect.

13 years agoTry to adhere to style(9) and be consistent within this file.
marius [Sat, 2 Oct 2010 17:20:30 +0000 (17:20 +0000)]
Try to adhere to style(9) and be consistent within this file.

13 years agoFix some KTR arguments that were breaking the LINT build.
nwhitehorn [Sat, 2 Oct 2010 17:19:38 +0000 (17:19 +0000)]
Fix some KTR arguments that were breaking the LINT build.

Pointy hat to: me

13 years agoRelease the vnode lock and close the linker file vnode earlier in
kib [Sat, 2 Oct 2010 16:04:50 +0000 (16:04 +0000)]
Release the vnode lock and close the linker file vnode earlier in
the linker_load_file methods. The change is that the consequent
linker_file_unload() call is not under the vnode lock anymore.
This prevents the LOR between kernel linker sx xlock and vnode lock,
because linker_file_unload() relocks kernel linker lock.

MFC after: 2 weeks

13 years agoCleanup some example and empty directories that were left around after
dim [Sat, 2 Oct 2010 11:45:18 +0000 (11:45 +0000)]
Cleanup some example and empty directories that were left around after
the last import of contrib/llvm.

Suggested by: obrien
Approved by: rpaulo (mentor)

13 years agoCosmetic: make it less confusing when displaying RAID 1 level, that might
kib [Sat, 2 Oct 2010 08:18:19 +0000 (08:18 +0000)]
Cosmetic: make it less confusing when displaying RAID 1 level, that might
be 1+0 as well.

PR: kern/150936
MFC after: 2 weeks

13 years agoFix up whatis/apropos issue displaying all output on a single due to
gordon [Sat, 2 Oct 2010 06:55:04 +0000 (06:55 +0000)]
Fix up whatis/apropos issue displaying all output on a single due to
forgotten quotes.

Submitted by: Brandon Gooch
Approved by: wes (mentor, implicit)

13 years agoo) Allow devices to override the MDIO read and write functions presented to
jmallett [Sat, 2 Oct 2010 05:43:17 +0000 (05:43 +0000)]
o) Allow devices to override the MDIO read and write functions presented to
   the miibus attached to octe interfaces.
o) Add an SMI/MDIO interface to the MV88E61XX and use it for the switch PHY on
   the Lanner MR-320.  An actual driver for the switch PHY will come later.
   Note that for now it intercepts and fakes MII_BMSR reads to prevent the
   miibus from talking to anything but the switch itself.

13 years agoRather than shifting offsets by three, set register offset to 3. All our
jmallett [Sat, 2 Oct 2010 05:38:45 +0000 (05:38 +0000)]
Rather than shifting offsets by three, set register offset to 3.  All our
bus interface does that's special here now is to use a 64-bit register size.
In theory, uart(4) ought to support a regsz as well as regshft and support
64-bit registers directly.

Also use the UART class's range rather than a hand-coded 1024 for the address
range.

13 years agoUse ABI-aware macros for setting up a fake frame.
jmallett [Sat, 2 Oct 2010 01:29:09 +0000 (01:29 +0000)]
Use ABI-aware macros for setting up a fake frame.

13 years agoRemove extra cpu setting and commented-out devices, some of which don't exist.
jmallett [Sat, 2 Oct 2010 01:28:18 +0000 (01:28 +0000)]
Remove extra cpu setting and commented-out devices, some of which don't exist.

13 years agoMake the OCTEON1 configuration look like a normal GENERIC configuration. In
jmallett [Sat, 2 Oct 2010 01:24:20 +0000 (01:24 +0000)]
Make the OCTEON1 configuration look like a normal GENERIC configuration.  In
particular, add PCI and USB buses and most devices.  Also move to using ULE,
adding INET6, turning on WITNESS by default, etc.

13 years agoUse __FBSDID() instead of RCSID() in most .S files under lib/msun/i386,
dim [Fri, 1 Oct 2010 20:14:36 +0000 (20:14 +0000)]
Use __FBSDID() instead of RCSID() in most .S files under lib/msun/i386,
and one under lib/msun/amd64.  This avoids adding the identifiers to the
.text section, and moves them to the .comment section instead.

Suggested by: bde
Approved by: rpaulo (mentor)

13 years agoMap the Open Firmware framebuffer console with write combining turned on,
nwhitehorn [Fri, 1 Oct 2010 19:02:31 +0000 (19:02 +0000)]
Map the Open Firmware framebuffer console with write combining turned on,
and set memory attributes appropriately for mmap() calls on /dev/console.
Xorg no longer uses /dev/console to mmap the framebuffer, so framebuffer
write combining support in X will arrive in the next patch.

13 years agoFix pmap_page_set_memattr() behavior in the presence of fictitious pages
nwhitehorn [Fri, 1 Oct 2010 18:59:30 +0000 (18:59 +0000)]
Fix pmap_page_set_memattr() behavior in the presence of fictitious pages
by just caching the mode for later use by pmap_enter(), following amd64.
While here, correct some mismerges from mmu_oea64 -> mmu_oea and clean
up some dead code found while fixing the fictitious page behavior.

13 years agoRemove extra semicolon.
yongari [Fri, 1 Oct 2010 17:51:55 +0000 (17:51 +0000)]
Remove extra semicolon.

13 years agoAllow write DMA to request larger DMA burst size to get better
yongari [Fri, 1 Oct 2010 17:46:15 +0000 (17:46 +0000)]
Allow write DMA to request larger DMA burst size to get better
performance on BCM5785.

Obtained from: OpenBSD

13 years agoput back the assigment to sched_time. It was correct, and
luigi [Fri, 1 Oct 2010 15:38:35 +0000 (15:38 +0000)]
put back the assigment to sched_time. It was correct, and
it was necessary.

Submitted by: Riccardo Panicucci

13 years agoOnly hide the ifa and not the tp under #ifdef INET as the tp is needed
bz [Fri, 1 Oct 2010 15:14:14 +0000 (15:14 +0000)]
Only hide the ifa and not the tp under #ifdef INET as the tp is needed
for locking evenwhen there is no INET.

MFC after: 3 days

13 years agoChange the output of mctest to give a summary of the results instead
gnn [Fri, 1 Oct 2010 14:36:36 +0000 (14:36 +0000)]
Change the output of mctest to give a summary of the results instead
of printing a long list.

Add a default base port, and default mulitcast address to the
runner script.

Add support for specifying a different local and remote interface
in the runner script.

MFC after: 1 week

13 years agolibc: Remove the i386 assembler version of strlen(3).
jilles [Fri, 1 Oct 2010 13:10:11 +0000 (13:10 +0000)]
libc: Remove the i386 assembler version of strlen(3).

On anything modern, the C version, which processes a word at a time, is much
faster. The Intel optimization manual explicitly warns against using REP
prefixes with SCAS or CMPS, which is exactly what the assembler version
does.

A simple test on a Phenom II showed the C version, compiled with -O2, to be
about twice as fast determining the length of 100000 strings between 0 and
255 bytes long.

MFC after: 2 weeks

13 years agoProper bracketing.
bz [Fri, 1 Oct 2010 11:48:14 +0000 (11:48 +0000)]
Proper bracketing.

PR: kern/151100
Submitted by: SunMinghao (sunminghao hotmail.com)
MFC after: 3 days

13 years ago- simplify by using one build script
mr [Fri, 1 Oct 2010 10:34:35 +0000 (10:34 +0000)]
- simplify by using one build script
- use label for mounting
- use GENERIC kernel
- create empty Pkg directory