dragonfly.git
11 years agoiso639: Sync with Library of Congress list.
Sascha Wildner [Fri, 7 Dec 2012 11:54:14 +0000 (12:54 +0100)]
iso639: Sync with Library of Congress list.

11 years agoRemove MFILES from kernel module Makefiles. It should not be needed.
Sascha Wildner [Thu, 6 Dec 2012 11:06:06 +0000 (12:06 +0100)]
Remove MFILES from kernel module Makefiles. It should not be needed.

11 years agokern.fwd.mk: Move comments to the start of the line.
Sascha Wildner [Thu, 6 Dec 2012 10:54:31 +0000 (11:54 +0100)]
kern.fwd.mk: Move comments to the start of the line.

Else they would get output to the screen.

Reported-by: tuxillo
11 years agoagp(4): Fix some minor issues.
Sascha Wildner [Thu, 6 Dec 2012 09:36:21 +0000 (10:36 +0100)]
agp(4): Fix some minor issues.

* Move AGP_DEBUG to the global 'options' file. No need to split it by
  platform.

* Add missing file to the Makefile.

* Fix a cast.

11 years agotwe(4): Sync with FreeBSD.
Sascha Wildner [Wed, 5 Dec 2012 22:28:33 +0000 (23:28 +0100)]
twe(4): Sync with FreeBSD.

Main change is making it MPSAFE. There's also some cleanup and misc
fixes.

I tested it with an Escalade 8506-8.

11 years agoRemove some unneeded semicolons across the tree.
Sascha Wildner [Wed, 5 Dec 2012 20:21:22 +0000 (21:21 +0100)]
Remove some unneeded semicolons across the tree.

11 years agokernel - Fix memory starvation issue w/tmpfs
Matthew Dillon [Wed, 5 Dec 2012 19:40:01 +0000 (11:40 -0800)]
kernel - Fix memory starvation issue w/tmpfs

* TMPFS relies on the pagedaemon to retire dirty pages to swap.  The normal
  buffer cache flushing won't do the job (nor do we want it to).  To avoid
  starving the system we change bio_page_alloc() to not dig into the
  system reserve when allocating pages for TMPFS.

Reported-by: tuxillo (Antonio Huete)
11 years agoagp: Fix a kernel panic on boot issue
François Tigeot [Wed, 5 Dec 2012 07:19:40 +0000 (08:19 +0100)]
agp: Fix a kernel panic on boot issue

* A pointer wasn't correctly initialized, leading to a
  Fatal trap 12: page fault while in kernel mode panic

* This commit fixes bug report #2467

Tested-by: Eric Christeson, David Shao
11 years agocluster - Stabilization
Matthew Dillon [Wed, 5 Dec 2012 07:32:30 +0000 (23:32 -0800)]
cluster - Stabilization

* Fix disconnect/reconnect sequence for autoconn (/etc/hammer2/autoconn).
  The pipe used to signal termination of the iocom_core() was not supposed
  to be closed by iocom_done().

* The shutdown code now simply sets DMSG_IOCOMF_EOF instead of trying to
  shutdown() the socket.

* Fix double mutex lock in dmsg_msg_alloc().

11 years agocluster - xdisk automatic BIO restart
Matthew Dillon [Wed, 5 Dec 2012 03:24:54 +0000 (19:24 -0800)]
cluster - xdisk automatic BIO restart

* The xdisk driver now detects failed transactions due to failed circuits
  and will restart the BIOs on another circuit or hold onto them until
  connectivity is restored and a new circuit is reforged.

  Automatic restarts only occur if the xa* disk is open()'d (i.e. mounted
  or being accessed by userland).  Kernel disk subsystem probes on attach
  will be failed normally and not stall on lost connectivity.

* subr_diskiocom now reports the correct DMSG error code for failed BIOs
  instead of reporting a kernel error code.

11 years agocluster - misc cleanup
Matthew Dillon [Wed, 5 Dec 2012 03:23:54 +0000 (19:23 -0800)]
cluster - misc cleanup

* Rename the routing function to something that I remember.

11 years agocluster - misc work
Matthew Dillon [Tue, 4 Dec 2012 22:30:42 +0000 (14:30 -0800)]
cluster - misc work

* Use a different API function for state-based reply in the volconf code.

11 years agocluster - Improve hammer2 connect directive, add /etc/hammer2/autoconn
Matthew Dillon [Tue, 4 Dec 2012 22:23:22 +0000 (14:23 -0800)]
cluster - Improve hammer2 connect directive, add /etc/hammer2/autoconn

* Improve the hammer2 connect directive, fixing a socket descriptor reuse
  race.

* Monitor the file /etc/hammer2/autoconn which contains a list of hosts
  the hammer2 service demon whould automatically maintain connections to.

* Adjust the dmsg_master_service() API to also allow an alternative
  signalling descriptor and callback to be specified.  This simplifies
  forced connection terminations.

* Fix descriptor leak in dmsg_master_service() on closure.

11 years agoFix LINT building.
Sascha Wildner [Tue, 4 Dec 2012 19:37:30 +0000 (20:37 +0100)]
Fix LINT building.

11 years agotwe(4): Remove non-existing flag from the Makefile.
Sascha Wildner [Tue, 4 Dec 2012 19:20:03 +0000 (20:20 +0100)]
twe(4): Remove non-existing flag from the Makefile.

11 years agokernel: Remove NULL checks after kmalloc(..., M_INTWAIT).
Sascha Wildner [Tue, 4 Dec 2012 18:43:47 +0000 (19:43 +0100)]
kernel: Remove NULL checks after kmalloc(..., M_INTWAIT).

M_INTWAIT will not return NULL unless M_NULLOK is passed too.

11 years agoagp: Fix a handful of issues with via agp support
rnoland [Mon, 21 Dec 2009 03:28:05 +0000 (03:28 +0000)]
agp: Fix a handful of issues with via agp support

  * Read the pci capability register to identify AGP 3 support
  * Add missing smaller aperture sizes for AGP3 chips.
  * Fix the aperture size calculation on AGP2 chips.
    All sizes between 32M and 256M reported as 256M.
  * Add \n to error string.

This all seems to get the CLE266 EPIA-M board agp working properly

Obtained-from: FreeBSD

11 years agoagp: Sync with FreeBSD
François Tigeot [Tue, 4 Dec 2012 07:16:33 +0000 (08:16 +0100)]
agp: Sync with FreeBSD

* Add support for AMD64 Family 10h processors. FreeBSD PR: kern/128331
* Add a few VIA bridges for Intel Core/Core 2 and VIA Nano processors
* Fix AGP debugging code
* Rename agp_x86_64.c to agp_amd64.c again, keeping the same file name
  on both DragonFly and FreeBSD
* Fix prototypes to be consistent
* offset parameter has vm_offset_t type
* vm_offset_t is unsigned and therefore can not be negative
* Use si_drv1 instead of dev2unit() inside agp(4).
* i810: Aperture size and stolen memory are now printed on one line
* Add pci ids for Intel Ironlake chipsets

11 years agocluster - Kernel cluster code work
Matthew Dillon [Tue, 4 Dec 2012 06:01:57 +0000 (22:01 -0800)]
cluster - Kernel cluster code work

* Major stability work.

* Misc bug fixes and cleanups.

* xdisk/diskiocom links are now relatively stable over a real network.
  Still need to deal with reconnects properly.

11 years agocluster - Add 'circ <msgid>' command to cluster debug shell
Matthew Dillon [Tue, 4 Dec 2012 06:00:37 +0000 (22:00 -0800)]
cluster - Add 'circ <msgid>' command to cluster debug shell

* Add a debug shell command to help test virtual circuits.

11 years agocluster - more libdmsg work
Matthew Dillon [Tue, 4 Dec 2012 05:58:16 +0000 (21:58 -0800)]
cluster - more libdmsg work

* Fix numerous bugs, including a nasty edge case in the encryption code.

* Do a better job locking things (state structures still need help).

11 years agokernel - Reorder dsched teardown to fix race
Matthew Dillon [Mon, 3 Dec 2012 18:20:29 +0000 (10:20 -0800)]
kernel - Reorder dsched teardown to fix race

* Remove the struct disk from its list prior to tearing down the device
  instead of after in order to fix a race related to disk enumerations.

* Wait for active references on the disk structure to go away before
  tearing down the device.

11 years agocluster - Major kernel component work (diskiocom, xdisk, kdmsg)
Matthew Dillon [Mon, 3 Dec 2012 05:40:13 +0000 (21:40 -0800)]
cluster - Major kernel component work (diskiocom, xdisk, kdmsg)

* Major implementation and stabilization work.  Fix lots of bugs,
  improve the AUTO flags and APIs.

* xdisk and diskiocom are now operational for the first time.

  xdisk is a consumer of remote block devices.  You pass it one end of
  a pipe and use an ioctl to create /dev/xa* devices.  The kernel's
  nominal disk management subsystem handles /dev/serno/* and will also
  probe the /dev/xa* devices to create the slices.

  diskiocom is a block device service.  An ioctl is used to connect a
  raw disk device to the network via a pipe.

* The hammer2 service demon ties the two together over the network.

* Initial testing with a local hammer2 service daemon looping da0 to xa0
  succeeded, ~46MB/sec unoptimized using dd.

11 years agocluster - circuit tracking, bug fixes
Matthew Dillon [Mon, 3 Dec 2012 05:37:55 +0000 (21:37 -0800)]
cluster - circuit tracking, bug fixes

* Misc virtual circuit tracking work and bug fixes

11 years agojme: Improve tiny packets transmission performance on low frequency CPU
Sepherosa Ziehau [Mon, 3 Dec 2012 01:53:20 +0000 (09:53 +0800)]
jme: Improve tiny packets transmission performance on low frequency CPU

Update TXCSR register a little bit often; mainly to improve timeliness of
packets transmission:
The TXCSR register is updated after certain amount of TX descriptors are
added to the hardware TX ring.  The default value of the amount of TX
descriptors are 16.  This value could be further tuned by per-device sysctl
node hw.jmeX.tx_wreg.

The default value improves tiny packets transmission performance w/ JMC250
on AMD970@2200Mhz (831Kpps -> 911Kpps), on AMD970@800Mhz (484Kpps -> 834Kpps)
and it does not increase CPU usage on AMD970@3500Mhz (CPU usage stays @26%,
JMC250 could only do 911Kpps).

11 years agobge: Improve tiny packets transmission performance on low frequency CPU
Sepherosa Ziehau [Sun, 2 Dec 2012 12:30:40 +0000 (20:30 +0800)]
bge: Improve tiny packets transmission performance on low frequency CPU

Update TX HOST_PROD register a little bit often; mainly to improve
timeliness of packets transmission:
The TX HOST_PROD register is updated after certain amount of TX
descriptors are added to the hardware TX ring.  The default value of the
amount of TX descriptors are 16.  This value could be further tuned by
per-device sysctl node hw.bgeX.tx_wreg.

The default value improves tiny packets transmission performance w/ 5721
on AMD970@800Mhz (770Kpps -> 800Kpps) and it does not increase CPU usage
on AMD970@3500Mhz (CPU usage stays @20%, 5721 could only do 810Kpps).

NOTE: Tuning hw.bgeX.tx_coal_bds to 32 could make packets transmission
performance w/ 5721 on AMD970@800Mhz to be 810Kpps, however, this shows
no performance improvement on commonly used CPU frequency like 2200Mhz
and 3500Mhz.  And this tuning could burden systems not using polling(4).

11 years agobge: Commit coalesce parameters change in polling code
Sepherosa Ziehau [Sun, 2 Dec 2012 10:42:37 +0000 (18:42 +0800)]
bge: Commit coalesce parameters change in polling code

11 years agomount_ntfs.8: Fix synopsis and add another example (from FreeBSD).
Sascha Wildner [Sun, 2 Dec 2012 23:36:26 +0000 (00:36 +0100)]
mount_ntfs.8: Fix synopsis and add another example (from FreeBSD).

11 years agokernel/ntfs: Remove two unused variables.
Sascha Wildner [Sun, 2 Dec 2012 17:55:41 +0000 (18:55 +0100)]
kernel/ntfs: Remove two unused variables.

11 years agokernel/ntfs: Add ntfs_statvfs.
Sascha Wildner [Sun, 2 Dec 2012 17:47:48 +0000 (18:47 +0100)]
kernel/ntfs: Add ntfs_statvfs.

This fixes for example df output for large filesystems.

11 years agoifconfig.8: Fix month
Sascha Wildner [Sun, 2 Dec 2012 10:22:38 +0000 (11:22 +0100)]
ifconfig.8: Fix month

11 years agomake upgrade: Remove /usr/include/bus/isa/isa_compat.h
Sascha Wildner [Sat, 1 Dec 2012 23:16:47 +0000 (00:16 +0100)]
make upgrade: Remove /usr/include/bus/isa/isa_compat.h

11 years agokernel/i386: With stl(4) and cy(4) made PCI only, remove COMPAT_OLDISA.
Sascha Wildner [Sat, 1 Dec 2012 23:09:09 +0000 (00:09 +0100)]
kernel/i386: With stl(4) and cy(4) made PCI only, remove COMPAT_OLDISA.

11 years agoUPDATING: Note that aic(4) was deISAized (forgot to mention at the time).
Sascha Wildner [Sat, 1 Dec 2012 22:54:54 +0000 (23:54 +0100)]
UPDATING: Note that aic(4) was deISAized (forgot to mention at the time).

11 years agocy(4): Remove ISA support.
Sascha Wildner [Sat, 1 Dec 2012 22:46:53 +0000 (23:46 +0100)]
cy(4): Remove ISA support.

11 years agokernel: Remove an #include I forgot previously.
Sascha Wildner [Sat, 1 Dec 2012 22:24:50 +0000 (23:24 +0100)]
kernel: Remove an #include I forgot previously.

11 years agokernel - Turn off machdep.pmap_mmu_optimize by default
Matthew Dillon [Sat, 1 Dec 2012 20:53:19 +0000 (12:53 -0800)]
kernel - Turn off machdep.pmap_mmu_optimize by default

* Turn off machdep.pmap_mmu_optimize by default.  There are some bugs that
  need to be resolved which are messing up both apache and rtorrent.

* (this feature is already turned off by default on RELEASE).

11 years agoMove stlstats(8) to usr.sbin/
Sascha Wildner [Sat, 1 Dec 2012 17:59:16 +0000 (18:59 +0100)]
Move stlstats(8) to usr.sbin/

11 years agoRemove stli(4). Also remove ISA support from stl(4).
Sascha Wildner [Sat, 1 Dec 2012 17:48:21 +0000 (18:48 +0100)]
Remove stli(4). Also remove ISA support from stl(4).

stli(4) was ISA only. stl(4) was for ISA/PCI controllers.

The stlload(8) and stlstty(8) utilities were for ISA controllers only,
so remove them too (along with ISA only firmware).

While here, clean up stl(4) a bit.

11 years agoUPDATING: Mention that bt(4) is PCI only now.
Sascha Wildner [Sat, 1 Dec 2012 12:44:28 +0000 (13:44 +0100)]
UPDATING: Mention that bt(4) is PCI only now.

11 years agoUPDATING: Use release version numbers.
Sascha Wildner [Sat, 1 Dec 2012 12:42:55 +0000 (13:42 +0100)]
UPDATING: Use release version numbers.

11 years agocluster - remote block device work
Matthew Dillon [Sat, 1 Dec 2012 10:24:49 +0000 (02:24 -0800)]
cluster - remote block device work

* xdisk(client) <-> diskiocom(server) now successfully does a full
  message transaction over a VC for xa_open().  Still lots to do.

* Fix I/O request deferrals since the open can occur before the VC is
  fully established.  This fixes a deadlock.

* Fix state->msg tracking in kern_dmsg.c.  Fixes a double-free.

11 years agohammer2 - cluster / hammer2 service daemon work
Matthew Dillon [Fri, 30 Nov 2012 23:43:28 +0000 (15:43 -0800)]
hammer2 - cluster / hammer2 service daemon work

* Reformulate the service demon to use the updated libdmsg API

* Add code to track remote block device SPANs and to interface to the
  local /dev/xdisk to create local block devices for the remotes.

11 years agohammer2 - cluster / libdmsg circuit work
Matthew Dillon [Fri, 30 Nov 2012 23:39:12 +0000 (15:39 -0800)]
hammer2 - cluster / libdmsg circuit work

* Major work on the virtual circuit code.  Note there are still some thread
  races.

* Major work on the spanning tree code to allow for future symmetric
  pathing support.

* fs_label is now part of the RB tree compare, along with pfs_fsid.
  (e.g. the serial number for BLK protocol services).

* Rip out the 'router' structure, replace with the 'circuit' structure.
  Embed circuit0 in the iocom.

11 years agohammer2 - reformualte cluster messaging interface
Matthew Dillon [Fri, 30 Nov 2012 23:37:17 +0000 (15:37 -0800)]
hammer2 - reformualte cluster messaging interface

* Reformulate the interface to use the revamped kern_dmsg.c API

11 years agokernel - Preliminary xdisk remote block driver for cluster
Matthew Dillon [Fri, 30 Nov 2012 23:35:44 +0000 (15:35 -0800)]
kernel - Preliminary xdisk remote block driver for cluster

* Basic infrastructure, ties into kern_dmsg.c cluster messaging system.
  (still a skeleton).

* Allows a userland program (aka hammer2 service) to add networked block
  devices to the local machine.

11 years agohammer2 - kernel cluster messaging support API work
Matthew Dillon [Fri, 30 Nov 2012 23:33:49 +0000 (15:33 -0800)]
hammer2 - kernel cluster messaging support API work

* Rework the API significantly

* Add circuit tracking and sniffing features

* Add flags to automate the LNK_CONN, LNK_SPAN, and LNK_CIRC state machines.

* Misc improvemenets.

11 years agokernel - Move dsched_disk_destroy_callback() call
Matthew Dillon [Fri, 30 Nov 2012 23:30:35 +0000 (15:30 -0800)]
kernel - Move dsched_disk_destroy_callback() call

* Move the dsched_disk_destroy_callback() call from the caller to the
  message handler to avoid racing a disk probe that might be running at
  the time (if a disk device is added and destroyed quickly).

11 years agokernel - Fix races in disk iteration and diskctx handling
Matthew Dillon [Fri, 30 Nov 2012 19:17:50 +0000 (11:17 -0800)]
kernel - Fix races in disk iteration and diskctx handling

* Add disk->d_refs to prevent a disk structure from being destroyed out
  from under an iteration.

* Redo the disk_enumeration() API to use markers and d_refs.

* Make adjustments to the dsched API.  In particular, do not return
  unreferenced tdio pointers in situations where they aren't used by
  the caller.

* Properly implement the ref count on the tdio's, one for each of the two
  lists the tdio belongs to, and ensure that dsched_thread_io_alloc()
  keeps an extra ref on the tdio after releasing the diskctx lock to prevent
  it from being ripped out while the code is pondering whether to place
  the tdio on the tdctx list.

* When deleting the tdio's for a tdctx try to destroy the diskctx.  That is,
  simply dereferencing it from the thread is not sufficient.

* When deleting the tdio's for a diskctx try to destroy the tdctx.  That is,
  simply dereferencing it from the diskctx is not sufficient.

* Handle destroy/ref races.

11 years agokernel: Fix typo (inavlid -> invalid).
Sascha Wildner [Fri, 30 Nov 2012 17:57:06 +0000 (18:57 +0100)]
kernel: Fix typo (inavlid -> invalid).

11 years agoint64_t is long on x86_64, and not always equivalent to long long.
Sascha Wildner [Thu, 29 Nov 2012 20:57:05 +0000 (21:57 +0100)]
int64_t is long on x86_64, and not always equivalent to long long.

Similarly, uint64_t is unsigned long.

For their min/max value definitions, change LL/ULL suffixes to L/UL.

The bug report has some C++ sample code that show issues resulting
from treating them as long long.

Reported-by: David Shao <davshao@gmail.com>
Dragonfly-bug: <http://bugs.dragonflybsd.org/issues/2461>

11 years agobt(4): Sync with FreeBSD.
Sascha Wildner [Thu, 29 Nov 2012 17:20:48 +0000 (18:20 +0100)]
bt(4): Sync with FreeBSD.

Main change is making it MPSAFE.

Taken-from: FreeBSD

11 years agobpf: Use bpf global token instead mplock to protect bpf stuffs
Sepherosa Ziehau [Thu, 29 Nov 2012 09:16:06 +0000 (17:16 +0800)]
bpf: Use bpf global token instead mplock to protect bpf stuffs

Submitted-by: tuxillo@ w/ modification
11 years agososendudp: Try to optimize out the additional mbuf alloc on output path
Sepherosa Ziehau [Thu, 29 Nov 2012 05:55:02 +0000 (13:55 +0800)]
sosendudp: Try to optimize out the additional mbuf alloc on output path

This optimization leaves enough space at the beginning of the mbuf, so
later on M_PREPEND() probably will not allocate addition mbuf.

This probably will not benefit any data that will be fragmented, e.g. by
IPv4, so this optimization is only performed when the size of data and
max size of protocol+link headers fit into one mbuf cluster.

This optimization could be turned off by net.inet.udp.sosend_prepend,
which is on by default.

11 years agobt(4): Remove support for ISA adapters.
Sascha Wildner [Thu, 29 Nov 2012 00:52:11 +0000 (01:52 +0100)]
bt(4): Remove support for ISA adapters.

The driver also supports PCI cards and we do want to keep this support,
because it is offered by current VirtualBox, for example.

11 years ago<sys/errno.h>: Try to fix errno's declaration a bit better for C++.
Sascha Wildner [Wed, 28 Nov 2012 23:26:41 +0000 (00:26 +0100)]
<sys/errno.h>: Try to fix errno's declaration a bit better for C++.

The test code from the bug ticket seems to work with it at least.

Reported-by: David Shao <davshao@gmail.com>
Dragonfly-bug:      <http://bugs.dragonflybsd.org/issues/2463>
In-discussion-with: vsrinivas

11 years agokernel - Fix proc_reparent() race/assertion panic
Matthew Dillon [Wed, 28 Nov 2012 18:21:38 +0000 (10:21 -0800)]
kernel - Fix proc_reparent() race/assertion panic

* Fix proc_reparent() race/assertion panic.  p_pptr changes can race,
  and the procedure had an assertion for the case.  Recode the procedure
  to retry on a mismatch instead of assert.

* Also move the old-parent-wakeup code into the procedure so it is
  properly executed in all cases.

Reported-by: Peter Avalos
11 years agomax_linkhdr: Increase from 16 to 20; mainly to cooperate SW VLAN tagging
Sepherosa Ziehau [Wed, 28 Nov 2012 09:21:53 +0000 (17:21 +0800)]
max_linkhdr: Increase from 16 to 20; mainly to cooperate SW VLAN tagging

11 years agoinstaller: Clarity language when asking the user for time-date info.
Matthew Dillon [Tue, 27 Nov 2012 21:50:54 +0000 (13:50 -0800)]
installer: Clarity language when asking the user for time-date info.

* Clarity language when asking the user for time-date info so the
  user is not confused between UTC vs his or her selected timezone.
  The program wants the user to enter the time and date in his or her
  selected timezone.

11 years agovkernel - Add machdep.wall_cmos_clock sysctl.
Antonio Huete Jimenez [Mon, 26 Nov 2012 23:14:14 +0000 (00:14 +0100)]
vkernel - Add machdep.wall_cmos_clock sysctl.

Programs that rely on this sysctl won't exit due its absence
and they will consider CMOS doesn't keep the local time.
Note the sysctl is read-only so programs like adjkerntz(8)
won't be able to change it.

11 years agoinstaller: Fix handling of CMOS set to UTC vs. CMOS set to wall time.
Sascha Wildner [Mon, 26 Nov 2012 18:39:57 +0000 (19:39 +0100)]
installer: Fix handling of CMOS set to UTC vs. CMOS set to wall time.

It was reversed in the installer (compared to tzsetup(8)).

/etc/wall_cmos_clock needs to be created when the CMOS is set to the
time of the clock on the wall. While here, also add removing the file
in case the user selects 'CMOS set to UTC'. This is if he chose the
other setting previously and then tried again.

Reported-by: many
Dragonfly-bug: <http://bugs.dragonflybsd.org/issues/39>
Dragonfly-bug: <http://bugs.dragonflybsd.org/issues/2060>

11 years agopolling: Increase default status polling fraction to 80
Sepherosa Ziehau [Mon, 26 Nov 2012 13:41:20 +0000 (21:41 +0800)]
polling: Increase default status polling fraction to 80

Before we bumped the polling frequency from 1000 to 2000, the status
polling fraction was 20; keep the original ratio.

11 years agobce: Commit coalesce parameters changes in polling code
Sepherosa Ziehau [Mon, 26 Nov 2012 13:38:48 +0000 (21:38 +0800)]
bce: Commit coalesce parameters changes in polling code

11 years agobce: Improve tiny packets transmission performance on low frequency CPU
Sepherosa Ziehau [Mon, 26 Nov 2012 13:32:42 +0000 (21:32 +0800)]
bce: Improve tiny packets transmission performance on low frequency CPU

Update TX HOST_IDX and HOST_BSEQ registers a little bit often; mainly to
improve timeliness of packets transmission:
The TX HOST_IDX and HOST_BSEQ registers are updated after certain amount
of TX descriptors are added to the hardware TX ring.  The default value
of the amount of TX descriptors are 8.  This value could be further
tuned by per-device sysctl node hw.bceX.tx_wreg.

The default value greatly improves tiny packets transmission performance
w/ 5709C on AMD970@800Mhz (513Kpps -> 1Mpps) and it does not increase CPU
usage on AMD970@3500Mhz (CPU usage stays @24%, 5709C could only do 1Mpps).

11 years agobce: Don't mess w/ MQ_COMMAND on TX path; it seems useless
Sepherosa Ziehau [Sun, 25 Nov 2012 12:57:57 +0000 (20:57 +0800)]
bce: Don't mess w/ MQ_COMMAND on TX path; it seems useless

Tested-with: 5706C, 5708C and 5709C

11 years agobce: Reformat comment in bce_npoll_compat a little bit
Sepherosa Ziehau [Sun, 25 Nov 2012 11:00:50 +0000 (19:00 +0800)]
bce: Reformat comment in bce_npoll_compat a little bit

11 years agoigb: Use SYSCTL_ADD_INT for hw.igbX.tx_wreg_nsegs
Sepherosa Ziehau [Sun, 25 Nov 2012 06:35:31 +0000 (14:35 +0800)]
igb: Use SYSCTL_ADD_INT for hw.igbX.tx_wreg_nsegs

It actually does not have any constraints.

11 years agoemx: Use SYSCTL_ADD_INT for hw.emxX.wreg_tx_nsegs
Sepherosa Ziehau [Sun, 25 Nov 2012 06:34:21 +0000 (14:34 +0800)]
emx: Use SYSCTL_ADD_INT for hw.emxX.wreg_tx_nsegs

It actually does not have any constraints.

11 years agobnx: Improve tiny packets transmission performance on low frequency CPU
Sepherosa Ziehau [Sun, 25 Nov 2012 06:08:31 +0000 (14:08 +0800)]
bnx: Improve tiny packets transmission performance on low frequency CPU

- Reduce TX coalesce BDs from 128 to 64.  This improves tiny packets
  transmission performance on AMD970@2200Mhz a little bit (1.40Mpps
  -> 1.48Mpps)
- Update TX PROD register a little bit often; mainly to improve timeliness
  of packets transmission:
  The TX PROD register is updated after certain amount of TX descriptors
  are added to the hardware TX ring.  The default value of the amount of
  TX descriptors are 8.  This value could be further tuned by per-device
  sysctl node hw.bnxX.tx_wreg.  The default value greatly improves tiny
  packets performance on AMD970@800Mhz (780Kpps -> 1.11Mpps) and it does
  not increase CPU usage on AMD970@3500Mhz (CPU usage stays @31%).

11 years agobnx: Commit coalesce parameters changes in polling code
Sepherosa Ziehau [Sat, 24 Nov 2012 13:24:14 +0000 (21:24 +0800)]
bnx: Commit coalesce parameters changes in polling code

11 years agokernel: Remove some code that does nothing.
Sascha Wildner [Sat, 24 Nov 2012 23:13:20 +0000 (00:13 +0100)]
kernel: Remove some code that does nothing.

11 years agoFix some typos (unknow -> unknown).
Sascha Wildner [Sat, 24 Nov 2012 22:14:45 +0000 (23:14 +0100)]
Fix some typos (unknow -> unknown).

11 years agousb4bsd: Initial round of build fixes for serial drivers.
Sascha Wildner [Sat, 24 Nov 2012 20:35:21 +0000 (21:35 +0100)]
usb4bsd: Initial round of build fixes for serial drivers.

Not hooked into the build because some things are still missing,
most importantly ucom(4) porting.

11 years agondis(4): Bring in some fixes from FreeBSD.
Sascha Wildner [Sat, 24 Nov 2012 20:11:30 +0000 (21:11 +0100)]
ndis(4): Bring in some fixes from FreeBSD.

11 years agondis(4): Make NDIS compile with usb4bsd too.
Sascha Wildner [Sat, 24 Nov 2012 20:22:55 +0000 (21:22 +0100)]
ndis(4): Make NDIS compile with usb4bsd too.

However, testing results are pretty much as modest as with the old USB
stack. Maybe someone else has more luck.

New files taken from FreeBSD.

11 years agousb4bsd: Fix buildkernel with device usb4bsd.
Sascha Wildner [Fri, 23 Nov 2012 18:46:19 +0000 (19:46 +0100)]
usb4bsd: Fix buildkernel with device usb4bsd.

11 years agowpa_supplicant.8: Document all options.
Sascha Wildner [Fri, 23 Nov 2012 18:11:29 +0000 (19:11 +0100)]
wpa_supplicant.8: Document all options.

Reported-by: Max Herrgard <herrgard@gmail.com>
Taken-from:  FreeBSD

11 years agowpa_supplicant.conf.5: Fix section order.
Sascha Wildner [Fri, 23 Nov 2012 18:02:53 +0000 (19:02 +0100)]
wpa_supplicant.conf.5: Fix section order.

11 years agoVFS quota: hpfs is now read-only, remove support
François Tigeot [Fri, 23 Nov 2012 06:57:26 +0000 (07:57 +0100)]
VFS quota: hpfs is now read-only, remove support

11 years agoDocumentation: unrhdr to idr conversion
François Tigeot [Mon, 19 Nov 2012 09:30:34 +0000 (10:30 +0100)]
Documentation: unrhdr to idr conversion

11 years agovkernel: Make "options IPX" compilable into the vkernel.
Sascha Wildner [Thu, 22 Nov 2012 19:44:23 +0000 (20:44 +0100)]
vkernel: Make "options IPX" compilable into the vkernel.

Usage of "errno" in userspace as a variable name results in sad code.
Rename them to "error" instead.

Reported-by: tuxillo
11 years agokernel: Remove some NULL checks after kmalloc(..., M_WAITOK).
Sascha Wildner [Thu, 22 Nov 2012 17:27:14 +0000 (18:27 +0100)]
kernel: Remove some NULL checks after kmalloc(..., M_WAITOK).

11 years agokernel: Use M_ZERO instead of bzero()ing.
Sascha Wildner [Thu, 22 Nov 2012 17:24:47 +0000 (18:24 +0100)]
kernel: Use M_ZERO instead of bzero()ing.

11 years agoah[cd](4): Clean up some #ifdefs so it's clearer what we are doing.
Sascha Wildner [Thu, 22 Nov 2012 16:41:11 +0000 (17:41 +0100)]
ah[cd](4): Clean up some #ifdefs so it's clearer what we are doing.

11 years agoemx: Don't update TDT upon very packets transmitted
Sepherosa Ziehau [Thu, 22 Nov 2012 12:53:36 +0000 (20:53 +0800)]
emx: Don't update TDT upon very packets transmitted

The TDT is updated only after certain amount of TX descriptors are
added to the hardware TX ring.  The default value of the amount of
TX descriptors are 8.  This value could be further tuned by per-device
sysctl node hw.emxX.tx_wreg_nsegs.

CPU usage when transmitting 18bytes tiny UDP datagrams @1.48Mpps with
82571 on AMD970 (3500Mhz) are reduced from 50% to 30% with the default
value.

Inspired-by: FreeBSD netmap's paper
11 years agoigb: Don't update TDT upon very packets transmitted
Sepherosa Ziehau [Thu, 22 Nov 2012 12:02:31 +0000 (20:02 +0800)]
igb: Don't update TDT upon very packets transmitted

The TDT is updated only after certain amount of TX descriptors are
added to the hardware TX ring.  The default value of the amount of
TX descriptors are 8.  This value could be further tuned by per-device
sysctl node hw.igbX.tx_wreg_nsegs.

CPU usage when transmitting 18bytes tiny UDP datagrams @1.48Mpps with
82575 on AMD970 (3500Mhz) are reduced from 55% to 38% with the default
value.

Inspired-by: FreeBSD netmap's paper
11 years agopktgen: Rework to improve performance using low CPU frequency
Sepherosa Ziehau [Sat, 17 Nov 2012 10:07:21 +0000 (18:07 +0800)]
pktgen: Rework to improve performance using low CPU frequency

The performance improvement is measured on AMD970 operating at 800Mhz
w/ Intel 82575 NIC using 18bytes UDP datagrams.

- Don't refill and allocate new packets for each transmission, instead
  a set of pregenerated packets are used.
- Use NIC driver's "txeof" to tick out more packets instead of bindly
  trying to inject more packets into NIC's TX ring.

These above two modifications give me ~200Kpps (710Kpps -> 910Kpps)
transmission rate improvement.

- Don't use hardware TX checksum offloading feature.  Since the set of
  packets to be used are pregenerated, this will not hurt performance,
  instead, this could improve performance on certain hardwares, e.g.
  Intel's NIC chipsets which require additional TX descriptor for setting
  up hardware TX checksum offload context.

The above modification give me additional ~40Kpps transmission rate
improvement.

11 years agohpfs - Fix a couple panics and a little cleanup.
Antonio Huete Jimenez [Thu, 22 Nov 2012 10:47:58 +0000 (11:47 +0100)]
hpfs - Fix a couple panics and a little cleanup.

* Fix compilation with HPFS_DEBUG.

* Fix a panic due CNP_PDIRUNLOCK flag not being cleared.

* Fix a panic where returned vnode after a lookup is not
  NULL in the ENOENT case.

* Disable write support completely. It was pretty minimal
  and operations like create or rename were not supported.

It has been tested with a filesystem created by OS/2 Warp 2.1.
Copying data out of it worked fine, but there is still an
outstanding issue with overlapping buffers.

11 years agokernel - Fix NFS panic when competing clients collide on hardlink
Matthew Dillon [Wed, 21 Nov 2012 17:59:55 +0000 (09:59 -0800)]
kernel - Fix NFS panic when competing clients collide on hardlink

* Fix recursive execlusive lock on vnode.

* For example, if several boxes are nfs booting and have the same writable
  /var/log, simultanious attempts to rotate logs can cause a panic due to
  higher layers of the OS's hardlink code not being able to detect
  duplicate vnodes which are created by some other client racing the same
  operation.

* Adds CNP_NOTVP and cn_notvp flag which vop_compat_nlink() uses to notify
  NFS's VOP_LOOKUP function to check for a vnode lookup collision.

-Matt

11 years agoRemove some duplicated semicolons (mostly in the kernel).
Sascha Wildner [Wed, 21 Nov 2012 17:13:43 +0000 (18:13 +0100)]
Remove some duplicated semicolons (mostly in the kernel).

11 years agohptiop(4): Add support for RocketRAID 4520 and 4522.
Sascha Wildner [Wed, 21 Nov 2012 03:26:47 +0000 (04:26 +0100)]
hptiop(4): Add support for RocketRAID 4520 and 4522.

This adds support for HighPoint's recent 6Gb/s SAS/SATA RAID cards.

It's not tested in DragonFly, unfortunately, due to lack of hardware,
but I'm confident. :)

Taken-from: FreeBSD

11 years agoigb: Dispatch callout to the CPU that is checking status
Sepherosa Ziehau [Tue, 20 Nov 2012 12:26:52 +0000 (20:26 +0800)]
igb: Dispatch callout to the CPU that is checking status

This reduces sporadic contentions on the main serializer.

11 years agoaac(4): Add a missing unlock.
Sascha Wildner [Tue, 20 Nov 2012 05:23:49 +0000 (06:23 +0100)]
aac(4): Add a missing unlock.

11 years agoiir(4): Clean up some unused code.
Sascha Wildner [Tue, 20 Nov 2012 04:32:38 +0000 (05:32 +0100)]
iir(4): Clean up some unused code.

11 years agoext2fs - Clear CNP_PDIRUNLOCK flag after a lookup.
Antonio Huete Jimenez [Mon, 19 Nov 2012 22:32:16 +0000 (23:32 +0100)]
ext2fs - Clear CNP_PDIRUNLOCK flag after a lookup.

vop_compat_nresolve() unlocks the vnode when flag CNP_PDIRUNLOCK
is set, so either we clear the flag or avoid unlocking the vnode
after an ext2_lookup().

DragonFly-bug: http://bugs.dragonflybsd.org/issues/2454

Reported by: Skylar Hawk

11 years agoUnbreak buildworld.
Sascha Wildner [Mon, 19 Nov 2012 17:56:42 +0000 (18:56 +0100)]
Unbreak buildworld.

Reported-by: marino
11 years ago/usr/Makefile: Add pkgsrc-create-shallow option
John Marino [Mon, 19 Nov 2012 17:43:52 +0000 (18:43 +0100)]
/usr/Makefile: Add pkgsrc-create-shallow option

11 years agomake distribution: Remove some symlinks before (re-)creating them.
Sascha Wildner [Mon, 19 Nov 2012 16:59:05 +0000 (17:59 +0100)]
make distribution: Remove some symlinks before (re-)creating them.

Reported-by: marino
11 years agonetstat: Make netstat -rn more resilient to having the routing table changing
Sepherosa Ziehau [Mon, 19 Nov 2012 09:51:24 +0000 (17:51 +0800)]
netstat: Make netstat -rn more resilient to having the routing table changing

Obtained-from: FreeBSD 176289 by jhb
  "Make netstat -rn more resilient to having the routing table change out from
   under it while running.  Note that this is still not perfect:
   - Try to do something intelligent if kvm_read() fails to read a routing
     table structure such as an rtentry, radix_node, or ifnet.
   - Don't follow left and right node pointers in radix_nodes unless
     RNF_ACTIVE is set in rn_flags.  This avoids walking through freed
     radix_nodes."

11 years agomchian: Sync w/ FreeBSD a little bit
Sepherosa Ziehau [Mon, 19 Nov 2012 09:38:44 +0000 (17:38 +0800)]
mchian: Sync w/ FreeBSD a little bit

subr_mchain.c CVS 1.{6, 8, 9, 10, 16, 18}

Submitted-by: Alexey Slynko w/ modification by me
DragonFly-bug: http://bugs.dragonflybsd.org/issues/80