dragonfly.git
16 years agoRemove fugly hack from the ATA channel interrupt handler. It never was really
Thomas E. Spanjaard [Mon, 29 Oct 2007 12:56:45 +0000 (12:56 +0000)]
Remove fugly hack from the ATA channel interrupt handler. It never was really
correct, and it has long become redundant.

16 years agoMaking dntpd server list match the recommendations at:
Justin C. Sherrill [Mon, 29 Oct 2007 12:51:21 +0000 (12:51 +0000)]
Making dntpd server list match the recommendations at:
http://www.pool.ntp.org/use.html

Dragonfly-bug: <http://bugs.dragonflybsd.org/issue832

16 years agorealloc_dynamic_table() and add_dyn_rule() may be called from interrupt thread
Sepherosa Ziehau [Mon, 29 Oct 2007 12:23:57 +0000 (12:23 +0000)]
realloc_dynamic_table() and add_dyn_rule() may be called from interrupt thread
and TCP/UDP thread, so instead of M_WAITOK, (M_INTWAIT | M_NULLOK) should be
used.

16 years agoComment out some "negative options" (options which make stuff not compile)
Sascha Wildner [Mon, 29 Oct 2007 10:31:27 +0000 (10:31 +0000)]
Comment out some "negative options" (options which make stuff not compile)
for more code coverage.

16 years ago- M_WAITOK will never return NULL
Sepherosa Ziehau [Mon, 29 Oct 2007 02:59:03 +0000 (02:59 +0000)]
- M_WAITOK will never return NULL
- White space

16 years agoNuke unnecessary NOMAN and .PATH
Sepherosa Ziehau [Mon, 29 Oct 2007 02:54:15 +0000 (02:54 +0000)]
Nuke unnecessary NOMAN and .PATH

16 years agoUse CRC_LEN consistently
Sepherosa Ziehau [Sun, 28 Oct 2007 02:29:07 +0000 (02:29 +0000)]
Use CRC_LEN consistently

16 years agoRemove two unused variables and model the copyright a bit more
Sascha Wildner [Sun, 28 Oct 2007 01:32:33 +0000 (01:32 +0000)]
Remove two unused variables and model the copyright a bit more
after /usr/src/COPYRIGHT.

16 years agoSwitch from libreadline to libedit.
Peter Avalos [Sat, 27 Oct 2007 22:36:47 +0000 (22:36 +0000)]
Switch from libreadline to libedit.

16 years agoSync with NetBSD.
Peter Avalos [Sat, 27 Oct 2007 22:27:16 +0000 (22:27 +0000)]
Sync with NetBSD.

16 years agoRemove OS name from the description.
Sascha Wildner [Sat, 27 Oct 2007 18:45:00 +0000 (18:45 +0000)]
Remove OS name from the description.

16 years agoAdjust column.
Sascha Wildner [Sat, 27 Oct 2007 18:25:39 +0000 (18:25 +0000)]
Adjust column.

16 years agodummynet_io() may be called from interrupt thread and TCP/UDP netisr thread,
Sepherosa Ziehau [Sat, 27 Oct 2007 12:19:43 +0000 (12:19 +0000)]
dummynet_io() may be called from interrupt thread and TCP/UDP netisr thread,
so instead of M_WAITOK, (M_INTWAIT | M_NULLOK) should be used when creating
flow queue.

16 years agoFix some mdoc issues.
Sascha Wildner [Fri, 26 Oct 2007 08:22:01 +0000 (08:22 +0000)]
Fix some mdoc issues.

Taken-from: FreeBSD

16 years agoWESTELL A90-200WG-01 works woth acx(4)
Sepherosa Ziehau [Thu, 25 Oct 2007 13:46:56 +0000 (13:46 +0000)]
WESTELL A90-200WG-01 works woth acx(4)

16 years ago- Switch dummynet(4) from mbuf based tagging to mtag.
Sepherosa Ziehau [Thu, 25 Oct 2007 13:13:18 +0000 (13:13 +0000)]
- Switch dummynet(4) from mbuf based tagging to mtag.
- Fix comment concerning mtag in sys/mbuf.h

Reviewed-by: dillon@
16 years agoIf answer to the repeated probe (same TTL as in the previous probe) packet
Hasso Tepper [Thu, 25 Oct 2007 08:12:42 +0000 (08:12 +0000)]
If answer to the repeated probe (same TTL as in the previous probe) packet
comes from different host (common in case of multipath routes), output it
to the new line - much better to read.

Obtained-from: FreeBSD with modifications

16 years agoUpdate for OpenSSL upgrade to 0.9.8g.
Peter Avalos [Thu, 25 Oct 2007 04:13:44 +0000 (04:13 +0000)]
Update for OpenSSL upgrade to 0.9.8g.

16 years agoMerge from vendor branch OPENSSL:
Peter Avalos [Thu, 25 Oct 2007 04:11:26 +0000 (04:11 +0000)]
Merge from vendor branch OPENSSL:
Import OpenSSL-0.9.8g.

16 years agoImport OpenSSL-0.9.8g.
Peter Avalos [Thu, 25 Oct 2007 04:11:26 +0000 (04:11 +0000)]
Import OpenSSL-0.9.8g.

16 years agoReactivate a vnode after associated it with deadfs after a forced unmount.
Matthew Dillon [Wed, 24 Oct 2007 21:56:41 +0000 (21:56 +0000)]
Reactivate a vnode after associated it with deadfs after a forced unmount.
This fixes numerous system panics that can occur due to the vnode's
unexpected change in state.

Submitted-by: "Nicolas Thery" <nthery@gmail.com>
16 years agoMove ipfw/dummynet from SI_SUB_PSEUDO to SI_SUB_PROTO_END, so they
Sepherosa Ziehau [Tue, 23 Oct 2007 15:01:15 +0000 (15:01 +0000)]
Move ipfw/dummynet from SI_SUB_PSEUDO to SI_SUB_PROTO_END, so they
could use services provided by netisr.

Submitted-by: Nicolas Thery <nthery@gmail.com>
16 years ago- Add jumbo buffer support for et(4), which supports up to 15794(*) bytes
Sepherosa Ziehau [Tue, 23 Oct 2007 14:28:42 +0000 (14:28 +0000)]
- Add jumbo buffer support for et(4), which supports up to 15794(*) bytes
  jumbo buffer size.  Allow jumbo buffer pool allocation to fail, if this
  happened, MTU still could be changed to a value less than 2026 bytes.
- Correct onboard memory configuration (*)
- Pack TX/RX enablement bool variable and newly added jumbo buffer supporting
  bool variable into a flag field in softc
- Increase TX spare segement count to reduce the possibility of busdma(9)
  operation failure
- Fix comment
- Replace magic number with macro in truephy, though the macros themselves
  are still quite magical
- Configure TX FIFO depth in truephy to support jumbo buffer

# (*) are based on experiments; peer is an Intel 82540EM, which supports
# 16218 bytes jumbo buffer size.

16 years agoRemove unnecessary argument.
Sascha Wildner [Tue, 23 Oct 2007 07:51:09 +0000 (07:51 +0000)]
Remove unnecessary argument.

16 years agoMerge ACPI resource manager changes From FreeBSD:
YONETANI Tomokazu [Tue, 23 Oct 2007 03:04:49 +0000 (03:04 +0000)]
Merge ACPI resource manager changes From FreeBSD:

acpi.c,v 1.158, acpivar.h,v 1.70
  Associate a device_t with an ACPI_HANDLE.  This make AcpiWalkNamespace more
  useful.  If ACPI-CA allowed null object handlers, we wouldn't need the
  placeholder function.

acpi.c,v 1.159-1.160 and others:
  Add support to ACPI to manage its own resources.  Previously, resource
  allocation was passed up to nexus.  Now, we probe sysresource objects and
  manage the resources they describe in a local rman pool.  This helps
  devices which attach/detach varying resources (like the _CST object) and
  module loads/unloads.  The allocation/release routines now check to see if
  the resource is described in a child sysresource object and if so,
  allocate from the local rman.  Sysresource objects add their resources to
  the pool and reserve them upon boot.  This means sysresources need to be
  probed before other ACPI devices.

  Changes include:
  * Add ordering to the child device probe.  The current order is:  system
  resource objects, embedded controllers, then everything else.
  * Make acpi_MatchHid take a handle instead of a device_t arg.
  * Replace acpi_{get,set}_resource with the generic equivalents.

acpi.c,v 1.175:
  Call device_identify routines after doing the namespace walk.  This is
  needed so that sysresource objects are created first to reserve all regions,
  then other devices can allocate from them.  Otherwise, acpi_timer (the only
  ACPI device with an identify routine), would allocate its resources from
  the nexus, causing the later sysresource reserve to fail.

  Debugging by:   Taku YAMAMOTO, Andrea Campi

And add ACPI attachment to some devices trying to obtain resources
from the ISA bus.

16 years agoFix more NULL structural dereferences that can occur when a process is in
Matthew Dillon [Sun, 21 Oct 2007 16:45:17 +0000 (16:45 +0000)]
Fix more NULL structural dereferences that can occur when a process is in
an exiting state.

Reported-by: Peter Avalos <pavalos@theshell.com>
16 years agoAdd a sysctl node, which could be used to turn off software TX power
Sepherosa Ziehau [Sun, 21 Oct 2007 09:50:33 +0000 (09:50 +0000)]
Add a sysctl node, which could be used to turn off software TX power
calibration.

16 years agoUse .Nm
Sascha Wildner [Sat, 20 Oct 2007 18:13:25 +0000 (18:13 +0000)]
Use .Nm

16 years agoUse .Nm
Sascha Wildner [Sat, 20 Oct 2007 17:56:47 +0000 (17:56 +0000)]
Use .Nm

16 years agoOops, missing space.
Sascha Wildner [Sat, 20 Oct 2007 13:43:00 +0000 (13:43 +0000)]
Oops, missing space.

16 years ago* Add an example how to create /etc/mtree style mtree(8) files.
Sascha Wildner [Sat, 20 Oct 2007 13:30:33 +0000 (13:30 +0000)]
* Add an example how to create /etc/mtree style mtree(8) files.

* Remove duplicate word.

Taken-from: FreeBSD

16 years ago- Change some comment's style
Sepherosa Ziehau [Sat, 20 Oct 2007 11:10:50 +0000 (11:10 +0000)]
- Change some comment's style
- White space changes
- Use POSIX int types

16 years ago- m_tag_get() expects MB_DONTWAIT not M_NOWAIT
Sepherosa Ziehau [Sat, 20 Oct 2007 10:28:44 +0000 (10:28 +0000)]
- m_tag_get() expects MB_DONTWAIT not M_NOWAIT
- Utilize MBTOM() in m_tag_alloc()

16 years agoM_WAITOK -> MB_WAIT
Sepherosa Ziehau [Sat, 20 Oct 2007 09:38:01 +0000 (09:38 +0000)]
M_WAITOK -> MB_WAIT

16 years ago- Instead of overriding the meaning of the embedded mbuf header's mh_nextpkt,
Sepherosa Ziehau [Sat, 20 Oct 2007 09:08:28 +0000 (09:08 +0000)]
- Instead of overriding the meaning of the embedded mbuf header's mh_nextpkt,
  add 'dn_next' field in dn_pkt, which is used to link dn_pkt into various
  packet queues; this reduces dependency on the embedded mbuf header
- Nuke DN_NEXT and DN_NEXT_NC, use dn_pkt.dn_next directly
- Nuke white space before ';'

16 years agoUse __offsetof()
Sepherosa Ziehau [Sat, 20 Oct 2007 07:36:20 +0000 (07:36 +0000)]
Use __offsetof()

16 years ago- Add macro to convert MTU to frame length
Sepherosa Ziehau [Sat, 20 Oct 2007 05:22:57 +0000 (05:22 +0000)]
- Add macro to convert MTU to frame length
- Strip CRC of the received frames
- Add macros for RX rings' buffer size

16 years agoAdd missing .Nm
Sascha Wildner [Fri, 19 Oct 2007 20:46:53 +0000 (20:46 +0000)]
Add missing .Nm

16 years agoAdd missing MLINKS.
Sascha Wildner [Fri, 19 Oct 2007 20:39:25 +0000 (20:39 +0000)]
Add missing MLINKS.

16 years agoRemove duplicate line and add missing MLINK.
Sascha Wildner [Fri, 19 Oct 2007 20:28:43 +0000 (20:28 +0000)]
Remove duplicate line and add missing MLINK.

16 years agoAdd missing comma.
Sascha Wildner [Fri, 19 Oct 2007 20:28:09 +0000 (20:28 +0000)]
Add missing comma.

16 years agoAdd some missing MLINKS and sort alphabetically.
Sascha Wildner [Fri, 19 Oct 2007 20:16:43 +0000 (20:16 +0000)]
Add some missing MLINKS and sort alphabetically.

16 years agoFix nrelease/Makefile from trying to execute whatever is in an environment
Thomas E. Spanjaard [Fri, 19 Oct 2007 18:02:22 +0000 (18:02 +0000)]
Fix nrelease/Makefile from trying to execute whatever is in an environment
variable which it never explicitly set before. ${ENV} ought to be ${ENVCMD},
which makes for weird errors if people use a Bourne shell like mksh with a
custom .shrc.

16 years agoRework bwi(4) LED support:
Sepherosa Ziehau [Fri, 19 Oct 2007 14:27:04 +0000 (14:27 +0000)]
Rework bwi(4) LED support:
- Reorder some macros and inline functions, so they can be used as needed
- Make three types of LED blinking work (references ath(4) soft LED)
- Customize Linksys LED behaviour
- Add two per-iface sysctl nodes for LED, update manpage accordingly

16 years agoAdd (proper) support for ATi SB600 southbridge SATA controllers in AHCI mode.
Thomas E. Spanjaard [Fri, 19 Oct 2007 11:53:14 +0000 (11:53 +0000)]
Add (proper) support for ATi SB600 southbridge SATA controllers in AHCI mode.
Also, adds a catch-all to the atapci ident routine for as of yet unidentified
AHCI SATA controllers.

Tested-by: Aggelos Economopoulos <aoiko@cc.ece.ntua.gr>

16 years agoBuffalo WLI-CB-G54S works well with bwi(4)
Sepherosa Ziehau [Thu, 18 Oct 2007 10:56:24 +0000 (10:56 +0000)]
Buffalo WLI-CB-G54S works well with bwi(4)

16 years agoFix a race between exit and kinfo_proc. proc->p_pgrp and the related
Matthew Dillon [Wed, 17 Oct 2007 23:30:31 +0000 (23:30 +0000)]
Fix a race between exit and kinfo_proc.  proc->p_pgrp and the related
session info can wind up being NULL if kinfo_proc catches the process while
it is trying to exit.

Reported-by: Peter Avalos <pavalos@theshell.com>
16 years agoEnabling TX/RX in et_init() will always fail when cable is not plugged in,
Sepherosa Ziehau [Wed, 17 Oct 2007 13:25:04 +0000 (13:25 +0000)]
Enabling TX/RX in et_init() will always fail when cable is not plugged in,
if this happens, we delay TX/RX enablement until link is up.

16 years agoMention that et(4) can receive oversized packets.
Sepherosa Ziehau [Wed, 17 Oct 2007 12:48:23 +0000 (12:48 +0000)]
Mention that et(4) can receive oversized packets.

Reminded-by: swildner@
16 years agoTurn on VLAN_MTU
Sepherosa Ziehau [Wed, 17 Oct 2007 12:01:57 +0000 (12:01 +0000)]
Turn on VLAN_MTU

16 years agoMake my Linksys WPC54G v3 LEDs work.
Sepherosa Ziehau [Wed, 17 Oct 2007 11:10:40 +0000 (11:10 +0000)]
Make my Linksys WPC54G v3 LEDs work.

It has two LEDs:
- "Power", which is marked as "always on" in SPROM
- "Link", which is marked as "activity" in SPROM

16 years agoAdd volume, super-cluster, cluster, and buffer abstractions to provide
Matthew Dillon [Tue, 16 Oct 2007 18:30:53 +0000 (18:30 +0000)]
Add volume, super-cluster, cluster, and buffer abstractions to provide
infrastructure for manipulating, reading, and writing various HAMMER
filesystem buffer types.

Properly set up allocation hints and use the A-list's as-of feature.

Completely implement super-cluster A-list layering and cluster/buffer
A-list layering.

(Preliminary) initialization of all primary on-disk structures and the
root directory.

16 years agoGive the A-list code the ability to do a forward or reverse allocation
Matthew Dillon [Tue, 16 Oct 2007 18:16:42 +0000 (18:16 +0000)]
Give the A-list code the ability to do a forward or reverse allocation
as-of a particular block.  This allows us to localize allocations.

Flesh out HAMMER's on-disk structures.

16 years agoReplace dummynet(4)'s callout by systimer, so you will not need to mess
Sepherosa Ziehau [Tue, 16 Oct 2007 11:28:40 +0000 (11:28 +0000)]
Replace dummynet(4)'s callout by systimer, so you will not need to mess
global HZ when using dummynet(4).

The timer frequency is controlled by net.inet.ip.dummynet.hz, which is
default to 1000 (originally recommended value for HZ); currently, it
should be changed only if dummynet(4) is not configured yet.

Reviewed-by: dillon@
16 years agoFix possible races between calling systimer callback fucntion and
Sepherosa Ziehau [Tue, 16 Oct 2007 11:12:59 +0000 (11:12 +0000)]
Fix possible races between calling systimer callback fucntion and
deleting a systimer.  Add comment about the fix.

Obtained-from: dillon@

16 years agoRemove hp300/m68k related junk.
Sascha Wildner [Tue, 16 Oct 2007 08:49:38 +0000 (08:49 +0000)]
Remove hp300/m68k related junk.

16 years agoFix SYNOPSIS.
Sascha Wildner [Sun, 14 Oct 2007 18:46:59 +0000 (18:46 +0000)]
Fix SYNOPSIS.

16 years agoAccording to RFC2711 routers shouldn't treat all packets with a Router
Hasso Tepper [Sun, 14 Oct 2007 18:15:19 +0000 (18:15 +0000)]
According to RFC2711 routers shouldn't treat all packets with a Router
Alert hop-by-hop option as 'ours', but examine this datagram more closely -
if the router isn't interested in particular protocol it should ignore the
packet and forward it as normal. So, treat packets as 'ours' if they are
for MLD only.

Reported-by: Andrew McDonald <andrew@mcdonald.org.uk>
Obtained-from: KAME

16 years agoAdd references to bwi(4).
Sascha Wildner [Sun, 14 Oct 2007 09:43:20 +0000 (09:43 +0000)]
Add references to bwi(4).

16 years agoAdd reference to wlan_ratectl(4) and fix copyright year.
Sascha Wildner [Sun, 14 Oct 2007 09:38:07 +0000 (09:38 +0000)]
Add reference to wlan_ratectl(4) and fix copyright year.

16 years agoMove assertion to the proper place, since the target CPU will always see
Sepherosa Ziehau [Sun, 14 Oct 2007 04:52:44 +0000 (04:52 +0000)]
Move assertion to the proper place, since the target CPU will always see
that the SYSTF_IPIRUNNING flag is set.

Approved-by: dillon@
16 years ago- Pull NetBSD David Young's bitops.h into sys/sys, nuke the local copies in
Sepherosa Ziehau [Sun, 14 Oct 2007 04:15:18 +0000 (04:15 +0000)]
- Pull NetBSD David Young's bitops.h into sys/sys, nuke the local copies in
  bwi(4), et(4) and rtw(4)
- Header inclusion cleanup

16 years agoAdd rcrun(8) reference.
Sascha Wildner [Sun, 14 Oct 2007 00:14:36 +0000 (00:14 +0000)]
Add rcrun(8) reference.

16 years agoRemove stale MLINK via make upgrade.
Sascha Wildner [Sun, 14 Oct 2007 00:03:10 +0000 (00:03 +0000)]
Remove stale MLINK via make upgrade.

16 years agoAdd a rcrun(8) manual page.
Sascha Wildner [Sat, 13 Oct 2007 23:51:32 +0000 (23:51 +0000)]
Add a rcrun(8) manual page.

16 years agoAdd some words about faststart. While here, sort references.
Sascha Wildner [Sat, 13 Oct 2007 23:39:46 +0000 (23:39 +0000)]
Add some words about faststart. While here, sort references.

16 years agoSave a copy of ethernet header instead of leaving it in the mbuf, so we
Sepherosa Ziehau [Sat, 13 Oct 2007 11:32:34 +0000 (11:32 +0000)]
Save a copy of ethernet header instead of leaving it in the mbuf, so we
will not get a dangling 'eh' pointer after calling some functions, which
changes the passed in mbuf (may free the part that 'eh' is pointed at).

This commit is intended to fix following bug:
http://leaf.dragonflybsd.org/mailarchive/bugs/2005-01/msg00056.html

The discussion about the fix began with:
http://leaf.dragonflybsd.org/mailarchive/bugs/2005-01/msg00147.html

16 years agoConstify ng_ether_input_p/ng_ether_input_orphan_p 'eh' parameter.
Sepherosa Ziehau [Sat, 13 Oct 2007 10:50:34 +0000 (10:50 +0000)]
Constify ng_ether_input_p/ng_ether_input_orphan_p 'eh' parameter.

16 years agoHook fec
Sepherosa Ziehau [Sat, 13 Oct 2007 10:45:39 +0000 (10:45 +0000)]
Hook fec

16 years agoShut up gcc41 warnings by clearing return values before doing any further
Sepherosa Ziehau [Sat, 13 Oct 2007 09:53:51 +0000 (09:53 +0000)]
Shut up gcc41 warnings by clearing return values before doing any further
processing in bpf_movein()

16 years agoConstify vlan_input_p 'eh' parameter.
Sepherosa Ziehau [Sat, 13 Oct 2007 09:43:19 +0000 (09:43 +0000)]
Constify vlan_input_p 'eh' parameter.

16 years agoSpelling fix: as follow -> as follows
Sascha Wildner [Sat, 13 Oct 2007 09:01:11 +0000 (09:01 +0000)]
Spelling fix: as follow -> as follows

16 years agoSpelling fix.
Peter Avalos [Sat, 13 Oct 2007 08:50:39 +0000 (08:50 +0000)]
Spelling fix.

16 years ago- Remove ef_input{EII,SNAP,8022}() 'eh' parameter, which is never used.
Sepherosa Ziehau [Sat, 13 Oct 2007 08:49:35 +0000 (08:49 +0000)]
- Remove ef_input{EII,SNAP,8022}() 'eh' parameter, which is never used.
- Constify ef_inputp 'eh' parameter.

16 years agoConstify 'eh', which will not be modified in ether_ipfw_chk() at all.
Sepherosa Ziehau [Sat, 13 Oct 2007 08:14:50 +0000 (08:14 +0000)]
Constify 'eh', which will not be modified in ether_ipfw_chk() at all.

16 years agoReduce code duplication: if_init() supressing is now done in ether_ioctl()
Sepherosa Ziehau [Sat, 13 Oct 2007 08:02:05 +0000 (08:02 +0000)]
Reduce code duplication: if_init() supressing is now done in ether_ioctl()

16 years agoMerge from vendor branch LESS:
Peter Avalos [Fri, 12 Oct 2007 20:05:50 +0000 (20:05 +0000)]
Merge from vendor branch LESS:
Import less-409.

Fix crash when viewing text with invalid UTF-8 sequences.

16 years agoImport less-409.
Peter Avalos [Fri, 12 Oct 2007 20:05:50 +0000 (20:05 +0000)]
Import less-409.

Fix crash when viewing text with invalid UTF-8 sequences.

16 years agoUpdate for OpenSSL 0.9.8f:
Peter Avalos [Fri, 12 Oct 2007 19:50:20 +0000 (19:50 +0000)]
Update for OpenSSL 0.9.8f:

Major changes between OpenSSL 0.9.8e and OpenSSL 0.9.8f:

      o Add gcc 4.2 support.
      o Add support for AES and SSE2 assembly lanugauge optimization
        for VC++ build.
      o Support for RFC4507bis and server name extensions if explicitly
        selected at compile time.
      o DTLS improvements.
      o RFC4507bis support.
      o TLS Extensions support.

Fixes CVE-2007-4995 and CVE-2007-5135. For more changes, see the CHANGES
file.

16 years agoImport OpenSSL-0.9.8f.
Peter Avalos [Fri, 12 Oct 2007 19:40:12 +0000 (19:40 +0000)]
Import OpenSSL-0.9.8f.

16 years agoMerge from vendor branch OPENSSL:
Peter Avalos [Fri, 12 Oct 2007 19:40:12 +0000 (19:40 +0000)]
Merge from vendor branch OPENSSL:
Import OpenSSL-0.9.8f.

16 years agoClean up the structural organization. Separate out A-lists and make
Matthew Dillon [Fri, 12 Oct 2007 18:57:45 +0000 (18:57 +0000)]
Clean up the structural organization.  Separate out A-lists and make
the header filenames conform to a single standard.  Add a super-cluster
header and reorganize the A-list topology used in the cluster header.

Augment A-lists to support hinting across A-list layers.  This allows
A-lists to be glued together hierarchically and to pass hinting information
between the layers.  This support will be used by HAMMER to manage cluster
allocation via the A-lists embedded in volume and super-cluster headers,
and to manage B-Tree, record, and data allocation via the A-lists embedded
in the cluster header and embedded in individual filesystem buffers.

16 years agoAdjust the description of HAMMER's storage limitations. I have rearranged
Matthew Dillon [Fri, 12 Oct 2007 17:17:20 +0000 (17:17 +0000)]
Adjust the description of HAMMER's storage limitations.  I have rearranged
the way clusters are organized in order to implement a layered storage
management mechanic within the cluster.  To do this the number of 16K
filesystem buffers in a cluster had to be limited to 4096.  This allows
several layered A-lists to be embedded in the cluster header instead of
just one but limits the cluster size to 64MB instead of 512M.  Without
any other changes this would have limited volumes to 2TB, which is too small.

Add a new super-cluster layer to support volumes larger then 2TB.  Instead
of being limited to 32768 clusters per volume we now have a limit of
16384 super-clusters per volume and 32768 clusters per super-cluster,
resulting in a limit of 32768TB per volume.

A super-cluster is just another 16K filesystem buffer which adds another
A-List layer to the volume's storage management scheme.  It turns out to
be totally trivial to implement.  Cluster allocations will still be optimal
in a super-cluster arrangement simply by caching the last super-cluster
an allocation was made from.

16 years agoAdd et(4), which supports Agere ET1310 based Ethernet chips (PCIe only)
Sepherosa Ziehau [Fri, 12 Oct 2007 14:12:42 +0000 (14:12 +0000)]
Add et(4), which supports Agere ET1310 based Ethernet chips (PCIe only)

  This chips supports two RX rings, one is currently used for packets
  whose size are smaller than 110 bytes, the other one is used for the
  rest packets sizes.

  Its RX interrupt moderation is quite similar to what bge(4) does:
  Two control variables, one is used to control how many packet should be
  received the other is used to control RX interrupt delay, RX interrupt
  moderation is achieved through the interaction of these two variables.

  Its TX interrupt moderation is more straightforward than RX's ;), you
  can tell hardware which TX segment should trigger interrupt.  It also
  has a hardware timer, which is set to 1Hz currently to prevent
  if_watchdog() from (mis)firing.

  I didn't figure out how to add polling(4) support for this chip, its
  RX state ring simply stops working if interrupts are disabled.  However,
  its hardware timer may be used to mimic polling(4) support.

  The missing features of the driver as of this commit:
  - Hardware checksum
  - Hardware vlan tagging
  - Jumbo buffer support
  Hopefully, they will be added later.

Add TruePHY (will any vendors name their PHY as FalsePHY one day?) support
into miibus(4) for Agere ET1011C PHY, which is used by et(4).  The data
sheet says model is 1 for ET1011C, while my testing card's model is 4; it
may be just a variant.

16 years agoregen
Sepherosa Ziehau [Fri, 12 Oct 2007 14:03:41 +0000 (14:03 +0000)]
regen

16 years ago- Add OUI for PHYs from Agere
Sepherosa Ziehau [Fri, 12 Oct 2007 14:02:55 +0000 (14:02 +0000)]
- Add OUI for PHYs from Agere
- Add Agere ET1011C id

16 years agoSimplify the code a lot - don't try to be too clever and handle chips with
Hasso Tepper [Fri, 12 Oct 2007 11:15:53 +0000 (11:15 +0000)]
Simplify the code a lot - don't try to be too clever and handle chips with
different number of ports in different way. Matching and attaching single
port chips in interface match/attach phase works just as fine as it already
did for multiport chips.

16 years agoSync usage() with arp.8
Thomas Nikolajsen [Thu, 11 Oct 2007 21:51:20 +0000 (21:51 +0000)]
Sync usage() with arp.8

16 years agoBring in some minor fixes from FreeBSD.
Sascha Wildner [Thu, 11 Oct 2007 10:22:46 +0000 (10:22 +0000)]
Bring in some minor fixes from FreeBSD.

16 years agoUpdate to tcpdump-3.9.8.
Peter Avalos [Thu, 11 Oct 2007 02:46:51 +0000 (02:46 +0000)]
Update to tcpdump-3.9.8.

Our pf includes are in <net/pf/>, so add appropriate patch file.

16 years agoMerge from vendor branch TCPDUMP:
Peter Avalos [Thu, 11 Oct 2007 02:37:58 +0000 (02:37 +0000)]
Merge from vendor branch TCPDUMP:
Import tcpdump-3.9.8:

September 10, 2007.  ken@xelerance.com.  Summary for 0.9.8 tcpdump release
Rework ARP printer
Rework OSPFv3 printer
Add support for Frame-Relay ARP
Decode DHCP Option 121 (RFC 3442 Classless Static Route)
Decode DHCP Option 249 (MS Classless Static Route) the same as Option 121
TLV: Add support for Juniper .pcap extensions
Print EGP header in new-world-order style
Converted print-isakmp.c to NETDISSECT
Moved AF specific stuff into af.h
Test subsystem now table driven, and saves outputs and diffs to one place
Require <net/pfvar.h> for pf definitions - allows reading of pflog formatted
libpcap files on an OS other than where the file was generated

16 years agoImport tcpdump-3.9.8:
Peter Avalos [Thu, 11 Oct 2007 02:37:58 +0000 (02:37 +0000)]
Import tcpdump-3.9.8:

September 10, 2007.  ken@xelerance.com.  Summary for 0.9.8 tcpdump release
Rework ARP printer
Rework OSPFv3 printer
Add support for Frame-Relay ARP
Decode DHCP Option 121 (RFC 3442 Classless Static Route)
Decode DHCP Option 249 (MS Classless Static Route) the same as Option 121
TLV: Add support for Juniper .pcap extensions
Print EGP header in new-world-order style
Converted print-isakmp.c to NETDISSECT
Moved AF specific stuff into af.h
Test subsystem now table driven, and saves outputs and diffs to one place
Require <net/pfvar.h> for pf definitions - allows reading of pflog formatted
libpcap files on an OS other than where the file was generated

16 years agoOur pf includes are in /usr/include/net/pf/, so adjust the appropriate
Peter Avalos [Thu, 11 Oct 2007 01:07:29 +0000 (01:07 +0000)]
Our pf includes are in /usr/include/net/pf/, so adjust the appropriate
files.

16 years agoResolve conflicts.
Peter Avalos [Thu, 11 Oct 2007 00:59:51 +0000 (00:59 +0000)]
Resolve conflicts.

16 years agoMerge from vendor branch LIBPCAP:
Peter Avalos [Thu, 11 Oct 2007 00:42:47 +0000 (00:42 +0000)]
Merge from vendor branch LIBPCAP:
Import libpcap-0.9.8:

Allow reading of pflog formatted libpcap files on an OS other than where
the file was generated.

16 years agoImport libpcap-0.9.8:
Peter Avalos [Thu, 11 Oct 2007 00:42:47 +0000 (00:42 +0000)]
Import libpcap-0.9.8:

Allow reading of pflog formatted libpcap files on an OS other than where
the file was generated.

16 years agoConsolidate local patches into contrib/, making updates easier.
Peter Avalos [Thu, 11 Oct 2007 00:21:59 +0000 (00:21 +0000)]
Consolidate local patches into contrib/, making updates easier.

16 years agoRemove the symlink, not the directory it points to.
Peter Avalos [Wed, 10 Oct 2007 23:51:13 +0000 (23:51 +0000)]
Remove the symlink, not the directory it points to.

16 years agoRemove the symlink, not the directory it points to.
Peter Avalos [Wed, 10 Oct 2007 23:03:22 +0000 (23:03 +0000)]
Remove the symlink, not the directory it points to.

16 years agoPrimary header file infrastructure and A-list implementation for the
Matthew Dillon [Wed, 10 Oct 2007 19:37:25 +0000 (19:37 +0000)]
Primary header file infrastructure and A-list implementation for the
HAMMER VFS.