]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years agoo Use NULL in pereference to 0 in pointer contexts.
imp [Wed, 11 Feb 2009 04:54:02 +0000 (04:54 +0000)]
o Use NULL in pereference to 0 in pointer contexts.
o Use newly minted KOBJMETHOD_END as appropriate
o fix prototype for root_setup_intr.

15 years agoo Introduce KOBJMETHOD_END for the end of the kobj list.
imp [Wed, 11 Feb 2009 04:52:46 +0000 (04:52 +0000)]
o Introduce KOBJMETHOD_END for the end of the kobj list.
o Fix a minor indentation problem.
o Put in the extra-strict KOBJMETHOD define, but commented out since
  the tree isn't yet ready.

Reviewed by: (1) was posted to arch@ without objection (and 1 go for it)

15 years agoAllocating 2 MSI messages do not seem to work on certain controllers
yongari [Wed, 11 Feb 2009 00:23:56 +0000 (00:23 +0000)]
Allocating 2 MSI messages do not seem to work on certain controllers
so use just 1 MSI message. This fixes regression introduced in
r188381.

Tested by: many

15 years agoAdded entries for Lava SP-PCI (1 serial + 1 parallel) PCI card. The
kaiw [Wed, 11 Feb 2009 00:08:03 +0000 (00:08 +0000)]
Added entries for Lava SP-PCI (1 serial + 1 parallel) PCI card. The
card is a multifunction PCI and report itself as two logical devices.

15 years agowhen operating as a slave, clear neighbor entries and the peer bss node
sam [Tue, 10 Feb 2009 23:52:28 +0000 (23:52 +0000)]
when operating as a slave, clear neighbor entries and the peer bss node
on beacon miss

while here, remove a stray pointless write of iv_bmiss_count

15 years agoclean neighbor entries on beacon miss
sam [Tue, 10 Feb 2009 23:51:18 +0000 (23:51 +0000)]
clean neighbor entries on beacon miss

15 years agodon't do phantom beacon miss checking for s/w beacon miss handling,
sam [Tue, 10 Feb 2009 23:48:29 +0000 (23:48 +0000)]
don't do phantom beacon miss checking for s/w beacon miss handling,
this can mistakenly drop events that cause the s/w bmiss timer to
never get re-armed

15 years agoCheck for device_set_devclass() errors and skip driver probe/attach if any.
mav [Tue, 10 Feb 2009 23:22:29 +0000 (23:22 +0000)]
Check for device_set_devclass() errors and skip driver probe/attach if any.
Attach call without devclass set crashes the system.

On resume AHCI driver sometimes tries to create duplicate adX device.
It is surely his own problem, but IMHO it is not a reason to crash here.
Other reasons are also possible.

15 years agoFix shutdown prototypes.
imp [Tue, 10 Feb 2009 23:17:20 +0000 (23:17 +0000)]
Fix shutdown prototypes.

15 years agoFix read_ivar prototype.
imp [Tue, 10 Feb 2009 23:13:48 +0000 (23:13 +0000)]
Fix read_ivar prototype.

15 years agoFix iicbus_intr, iicbus_write and device_read_ivar prototypes...
imp [Tue, 10 Feb 2009 22:50:23 +0000 (22:50 +0000)]
Fix iicbus_intr, iicbus_write and device_read_ivar prototypes...

15 years agofix arcmsr_attach, probe, detach, and shutdown prototypes to match newbus.
imp [Tue, 10 Feb 2009 22:46:36 +0000 (22:46 +0000)]
fix arcmsr_attach, probe, detach, and shutdown prototypes to match newbus.

15 years agoAdding dynamic sysctls no longer requires Giant.
jhb [Tue, 10 Feb 2009 22:39:44 +0000 (22:39 +0000)]
Adding dynamic sysctls no longer requires Giant.

Submitted by: rdivacky

15 years agoDon't reset the PHY probe retry counter within the loop so
marius [Tue, 10 Feb 2009 21:54:23 +0000 (21:54 +0000)]
Don't reset the PHY probe retry counter within the loop so
it will eventually terminate as intended.

Submitted by: Helen Zhang

15 years agoImprove r185008 so the streaming cache is only flushed when
marius [Tue, 10 Feb 2009 21:51:33 +0000 (21:51 +0000)]
Improve r185008 so the streaming cache is only flushed when
a mapping actually met the threshold.

15 years ago- Use the generally more appropriate PROM base rather than the
marius [Tue, 10 Feb 2009 21:48:42 +0000 (21:48 +0000)]
- Use the generally more appropriate PROM base rather than the
  kernel one as the non-faulting flush address in the loader so
  we can can change KERNBASE and VM_MIN_KERNEL_ADDRESS if we
  ever want to without needing to worry about using a compatible
  loader.
- Correctly check for LOADER_DEBUG.
- Add a missing const for page_sizes[].

15 years agoMark the BSP as being awake. This supresses the message
marcel [Tue, 10 Feb 2009 20:29:57 +0000 (20:29 +0000)]
Mark the BSP as being awake. This supresses the message
that not all usable CPUs could be woken up...

15 years agoUpdate comment. soalloc() is no longer performing M_WAITOK memory allocations.
emax [Tue, 10 Feb 2009 20:27:05 +0000 (20:27 +0000)]
Update comment. soalloc() is no longer performing M_WAITOK memory allocations.

Submitted by: ru
MFC after: 3 days

15 years agomark the CLR key installed for open auth stations such that it is reclaimed
sam [Tue, 10 Feb 2009 19:27:50 +0000 (19:27 +0000)]
mark the CLR key installed for open auth stations such that it is reclaimed
when net80211 tears down station state; without this we leak keycache slots

15 years agoadd hw.ath.bstuck to control the stuck beacon threshold
sam [Tue, 10 Feb 2009 19:26:42 +0000 (19:26 +0000)]
add hw.ath.bstuck to control the stuck beacon threshold

15 years agoon resume ah_curchan may be NULL if no channel change has been done;
sam [Tue, 10 Feb 2009 19:25:11 +0000 (19:25 +0000)]
on resume ah_curchan may be NULL if no channel change has been done;
workaround this by passing net80211's channel as we know it'll never
be null

Submitted by: trasz

15 years agoconsolidate conditional code
sam [Tue, 10 Feb 2009 19:23:25 +0000 (19:23 +0000)]
consolidate conditional code

15 years agoScanning all the formats for binary translation of modules loading can
attilio [Tue, 10 Feb 2009 15:50:19 +0000 (15:50 +0000)]
Scanning all the formats for binary translation of modules loading can
result in errors for a format loading but subsequent correct recognizing
for another format.

File format loading functions should avoid printing any additional
informations but just returning appropriate (and different between each
other) error condition, characterizing different informations.
Additively, the linker should handle appropriately different format
loading errors.

While a general mechanism is desired, fix a simple and common case on
amd64: file type is not recognized for link elf and confuses the linker.
Printout an error if all the registered linker classes can't recognize
and load the module.

Reviewed by: jhb
Sponsored by: Sandvine Incorporated

15 years agoTurn off the parallel make options accidentally turned on by the previous
kensmith [Tue, 10 Feb 2009 15:11:26 +0000 (15:11 +0000)]
Turn off the parallel make options accidentally turned on by the previous
commit.  Sigh.

Noticed by: rdivacky

15 years agoDue to some recent-ish work on GEOM/G_PART the c partition (by tradition
kensmith [Tue, 10 Feb 2009 15:07:57 +0000 (15:07 +0000)]
Due to some recent-ish work on GEOM/G_PART the c partition (by tradition
the whole disk) isn't available any more since it was redundant.  Just
use /dev/md0 instead of /dev/md0c to build the filesystem on.

Consulted-with: marcel

15 years agoTurn off the building of boot floppies for amd64/i386.
kensmith [Tue, 10 Feb 2009 15:03:08 +0000 (15:03 +0000)]
Turn off the building of boot floppies for amd64/i386.

15 years agoEliminate an unused definition.
alc [Tue, 10 Feb 2009 06:08:28 +0000 (06:08 +0000)]
Eliminate an unused definition.

15 years agoFix g_part_dumpconf and g_part_name prototpyes.
imp [Tue, 10 Feb 2009 02:43:07 +0000 (02:43 +0000)]
Fix g_part_dumpconf and g_part_name prototpyes.

Submitted by: marcel@

15 years agocorrectly fixup iv_des_chan after installing a new channel table; if the
sam [Tue, 10 Feb 2009 01:37:55 +0000 (01:37 +0000)]
correctly fixup iv_des_chan after installing a new channel table; if the
previous channel is now invalid we must set it to IEEE80211_CHAN_ANYC

15 years agoAdd option GEOM_PART_EBR by default on amd64 and i386.
marcel [Tue, 10 Feb 2009 00:08:39 +0000 (00:08 +0000)]
Add option GEOM_PART_EBR by default on amd64 and i386.

15 years agoHmm... "base? base : 0" and "base" should be equivalent.
cognet [Mon, 9 Feb 2009 22:59:22 +0000 (22:59 +0000)]
Hmm... "base? base : 0" and "base" should be equivalent.

Pointy hat to: cognet

MFC after: 3 days

15 years agoHook uslcom2 up to the build.
thompsa [Mon, 9 Feb 2009 22:38:55 +0000 (22:38 +0000)]
Hook uslcom2 up to the build.

15 years agoRemove usb2_config_td.[ch], they are no longer used.
thompsa [Mon, 9 Feb 2009 22:23:01 +0000 (22:23 +0000)]
Remove usb2_config_td.[ch], they are no longer used.

15 years agoMFp4 //depot/projects/usb; 157429, 157433, 157443
thompsa [Mon, 9 Feb 2009 22:18:11 +0000 (22:18 +0000)]
MFp4 //depot/projects/usb; 157429, 157433, 157443

Various changes wrt. usb procs and tasks.

Submitted by: Hans Petter Selasky

15 years agoCorrect sources for r188417
thompsa [Mon, 9 Feb 2009 22:14:38 +0000 (22:14 +0000)]
Correct sources for r188417

MFp4 //depot/projects/usb; 157412

Sync from svn.freebsd.org/base/user/thompsa/usb which is a minimal changeset
from oldUSB (no config_td).

This excludes the taskqueue changes (for the moment) as requested.

15 years agoMFp4 //depot/projects/usb; 157412
thompsa [Mon, 9 Feb 2009 22:12:47 +0000 (22:12 +0000)]
MFp4 //depot/projects/usb; 157412

Sync from svn.freebsd.org/base/user/thompsa/usb which is a minimal changeset
from oldUSB (no config_td).

This excludes the taskqueue changes (for the moment) as requested.

15 years agoMFp4 //depot/projects/usb; 157069, 157255
thompsa [Mon, 9 Feb 2009 22:11:02 +0000 (22:11 +0000)]
MFp4 //depot/projects/usb; 157069, 157255

 - Change "usb2_pause_mtx" so that it takes the timeout value in ticks
 - Make sure that attach waits for the generic probe to leave room for firware
   loader drivers and device specific USB Audio drivers like USB phone adapters

Submitted by: Hans Petter Selasky

15 years agoMFp4 //depot/projects/usb; 157203
thompsa [Mon, 9 Feb 2009 22:09:21 +0000 (22:09 +0000)]
MFp4 //depot/projects/usb; 157203

 - Remove CAM layer quirks

Submitted by: Hans Petter Selasky

15 years agoMFp4 //depot/projects/usb; 157069
thompsa [Mon, 9 Feb 2009 22:06:52 +0000 (22:06 +0000)]
MFp4 //depot/projects/usb; 157069

 - Change "usb2_pause_mtx" so that it takes the timeout value in ticks

Submitted by: Hans Petter Selasky

15 years agoMFp4 //depot/projects/usb; 157069, 157136, 157143, 157145, 157429
thompsa [Mon, 9 Feb 2009 22:05:25 +0000 (22:05 +0000)]
MFp4 //depot/projects/usb; 157069, 157136, 157143, 157145, 157429

 - USB serial drivers cleanup, factor out code
 - Simplify line state programming
 - Integrate uslcom from old USB stack

Submitted by: Hans Petter Selasky

15 years agoMFp4 //depot/projects/usb; 157100
thompsa [Mon, 9 Feb 2009 22:02:38 +0000 (22:02 +0000)]
MFp4 //depot/projects/usb; 157100

1. Move most of the ifnet logic into the usb2_ethernet module, this includes,
 - make all usb ethernet interfaces named ue%d
 - handle all threading in usb2_ethernet
 - provide default ioctl handler
 - handle mbuf rx
 - provide locked callbacks for init,start,stop,etc

2. Cleanup CDC-Ethernet driver.

Submitted by: Hans Petter Selasky
Obtained from: svn.freebsd.org/base/user/thompsa/usb [1]

15 years agoMFp4 //depot/projects/usb; 157069, 157429, 157430
thompsa [Mon, 9 Feb 2009 21:56:33 +0000 (21:56 +0000)]
MFp4 //depot/projects/usb; 157069, 157429, 157430

 - Change "usb2_pause_mtx" so that it takes the timeout value in ticks
 - Factor out USB ethernet and USB serial driver specific control request.
 - USB process naming cleanup.

Submitted by: Hans Petter Selasky

15 years agoMFp4 //depot/projects/usb
thompsa [Mon, 9 Feb 2009 21:50:04 +0000 (21:50 +0000)]
MFp4 //depot/projects/usb

 - Bring in ehci2_ixp4xx.c and ehci2_mbus.c from dev/usb.

Submitted by: Hans Petter Selasky

15 years agoMFp4 //depot/projects/usb; 157069,157296,157375,157384,157430
thompsa [Mon, 9 Feb 2009 21:47:39 +0000 (21:47 +0000)]
MFp4 //depot/projects/usb; 157069,157296,157375,157384,157430

 - Change "usb2_pause_mtx" so that it takes the timeout value in ticks
 - USB controller: EHCI High Speed Interrupt endpoint fix.
 - Fix OHCI and EHCI counting bug when multiple TD's are involved in
   a short USB transfer and a short packet happens on the non-last TD in the
   USB transfer frame.
 - USB process naming cleanup.

Submitted by: Hans Petter Selasky

15 years ago- Sanitize disk parameters retrieved from GEOM, as they are not guaranteed to
lulf [Mon, 9 Feb 2009 21:34:06 +0000 (21:34 +0000)]
- Sanitize disk parameters retrieved from GEOM, as they are not guaranteed to
  have sane values. It caused sysinstall to crash when installing on certain SD
  cards.

Discussed with: marcel

15 years agoMark udf(4) MPSAFE and add support for shared vnode locks during pathname
jhb [Mon, 9 Feb 2009 20:50:23 +0000 (20:50 +0000)]
Mark udf(4) MPSAFE and add support for shared vnode locks during pathname
lookups:
- Honor the caller's locking flags in udf_root() and udf_vget().
- Set VV_ROOT for the root vnode in udf_vget() instead of only doing it in
  udf_root().
- Honor the requested locking flags during pathname lookups in udf_lookup().
- Release the buffer holding the directory data before looking up the vnode
  for a given file to avoid a LOR between the "udf" vnode locks and
  "bufwait".
- Use vn_vget_ino() to handle ".." lookups.
- Special case "." lookups instead of calling udf_vget().  We have to do
  extra checking for the vnode lock for "." lookups.

15 years agoUse the same style as the rest of the file for the optional data string
jhb [Mon, 9 Feb 2009 20:42:51 +0000 (20:42 +0000)]
Use the same style as the rest of the file for the optional data string
after each path component rather than a GCC-ism.

15 years ago- Fix an issue where file attributes were not installed correctly during a Touch
lulf [Mon, 9 Feb 2009 20:13:55 +0000 (20:13 +0000)]
- Fix an issue where file attributes were not installed correctly during a Touch
  and SetAttrs operation.
- SetAttrs and Touch were incorrectly switched.

15 years agoRemove extra 'comma = 0' in socket state printing code, which otherwise
rwatson [Mon, 9 Feb 2009 18:19:58 +0000 (18:19 +0000)]
Remove extra 'comma = 0' in socket state printing code, which otherwise
could lead to an extra comma in output.

Submitted by:   Christoph Mallon <christoph dot mallon at gmx dot de>

15 years agoThe bounce zone sees its page number increased if multiple dma maps use it in
cognet [Mon, 9 Feb 2009 18:03:31 +0000 (18:03 +0000)]
The bounce zone sees its page number increased if multiple dma maps use it in
the same dma tag. However, it can happen multiple dma tags share the same
bounce zone too, so add a per-bounce zone map counter, and check it instead of
the dma tag map counter, to know if we have to alloc more pages.

Reported by: miwi
Reviewed by: scottl

15 years agoAdd padding to then end of the xsctp_xxx structures to
rrs [Mon, 9 Feb 2009 17:37:17 +0000 (17:37 +0000)]
Add padding to then end of the xsctp_xxx structures to
allow future changes to be able to maintain ABI compatibility

15 years agotulip_busdma_cleanup: pass correct vaddr (2nd arg) to bus_dmamem_free()
fjoe [Mon, 9 Feb 2009 17:07:29 +0000 (17:07 +0000)]
tulip_busdma_cleanup: pass correct vaddr (2nd arg) to bus_dmamem_free()

Found with: Coverity Prevent(tm)
CID: 1998

15 years agocam_periph_alloc: fix "invalid periph name" error condition
fjoe [Mon, 9 Feb 2009 17:02:54 +0000 (17:02 +0000)]
cam_periph_alloc: fix "invalid periph name" error condition

Found with: Coverity Prevent(tm)
CID: 130

15 years agoDo not self-initialize a variable.
fjoe [Mon, 9 Feb 2009 16:58:18 +0000 (16:58 +0000)]
Do not self-initialize a variable.

Found with: Coverity Prevent(tm)
CID: 3864, 3865

15 years agoRemove unused variable.
fjoe [Mon, 9 Feb 2009 16:57:07 +0000 (16:57 +0000)]
Remove unused variable.

Found with: Coverity Prevent(tm)
CID: 3693

15 years agoDestroy TX tag outside of loop scope.
fjoe [Mon, 9 Feb 2009 16:53:37 +0000 (16:53 +0000)]
Destroy TX tag outside of loop scope.

Found with: Coverity Prevent(tm)
CID: 3886

15 years agoProperly implement GIO_ATTR and CONS_GETINFO.
ed [Mon, 9 Feb 2009 15:55:21 +0000 (15:55 +0000)]
Properly implement GIO_ATTR and CONS_GETINFO.

It seems I didn't implement these two ioctl()'s properly, which meant
vidcontrol couldn't properly obtain certain terminal parameters.

15 years agos/SS_FDREF/SS_NOFDREF/
mbr [Mon, 9 Feb 2009 13:29:01 +0000 (13:29 +0000)]
s/SS_FDREF/SS_NOFDREF/

15 years agoFix minor spacing problem found by s9indent from last
rrs [Mon, 9 Feb 2009 11:42:23 +0000 (11:42 +0000)]
Fix minor spacing problem found by s9indent from last
commit.

15 years agoFix INET only build breakage with SCTP - pointy hat to me :-)
rrs [Mon, 9 Feb 2009 11:41:54 +0000 (11:41 +0000)]
Fix INET only build breakage with SCTP - pointy hat to me :-)

15 years agoComment out the assertion from r188321. It is not valid for nfs.
kib [Mon, 9 Feb 2009 11:32:23 +0000 (11:32 +0000)]
Comment out the assertion from r188321. It is not valid for nfs.

Reported by: alc

15 years agoRemove a stale comment from the clists code.
ed [Mon, 9 Feb 2009 11:27:56 +0000 (11:27 +0000)]
Remove a stale comment from the clists code.

We don't support quote bits.

15 years agoAvoid some cases of unnecessary page queues locking by vm_fault's delete-
alc [Mon, 9 Feb 2009 06:23:21 +0000 (06:23 +0000)]
Avoid some cases of unnecessary page queues locking by vm_fault's delete-
behind heuristic.

15 years agoDocument loader tunable hw.re.msi_disable.
yongari [Mon, 9 Feb 2009 05:09:52 +0000 (05:09 +0000)]
Document loader tunable hw.re.msi_disable.
Bump .Dd

15 years agoReclaim transmitted frames in re_tick(). This is for PCIe
yongari [Mon, 9 Feb 2009 04:59:13 +0000 (04:59 +0000)]
Reclaim transmitted frames in re_tick(). This is for PCIe
controllers that lose Tx completion interrupts under certain
conditions. With this change it's safe to use MSI on PCIe
controllers so enable MSI on these controllers.

15 years agomark M_LASTFRAG at the last fragment.
weongyo [Mon, 9 Feb 2009 04:39:16 +0000 (04:39 +0000)]
mark M_LASTFRAG at the last fragment.

Reviewed by: sam
MFC after: 3 weeks

15 years agoAdd the EBR scheme. The EBR scheme supports the Extended Boot Records
marcel [Sun, 8 Feb 2009 23:51:44 +0000 (23:51 +0000)]
Add the EBR scheme. The EBR scheme supports the Extended Boot Records
found inside extended partitions and used to create logical partitions.
At this time write/modify support is not (yet) present.
The EBR and MBR schemes both check the parent scheme. The MBR will
back-off when nested under another MBR, whereas the EBR only nests
under a MBR.

15 years ago- Fix in_cksum for big-endian MIPS: use correct compile-time check.
gonzo [Sun, 8 Feb 2009 23:43:36 +0000 (23:43 +0000)]
- Fix in_cksum for big-endian MIPS: use correct compile-time check.

15 years agoAllow gpe_offset to be set by the scheme. When gpe_offset is zero,
marcel [Sun, 8 Feb 2009 23:39:30 +0000 (23:39 +0000)]
Allow gpe_offset to be set by the scheme. When gpe_offset is zero,
or invalid, initialize it to the start of the partition. Adjust
the mediasize when the offset lies somewhere inside the partition.

15 years agoWhen bouncing pages, allow a new option to preserve the intra-page
imp [Sun, 8 Feb 2009 22:54:58 +0000 (22:54 +0000)]
When bouncing pages, allow a new option to preserve the intra-page
offset.  This is needed for the ehci hardware buffer rings that assume
this behavior.

This is an interim solution, and a more general one is being worked
on.  This solution doesn't break anything that doesn't ask for it
directly.  The mbuf and uio variants with this flag likely don't work
and haven't been tested.

Universe builds with these changes.  I don't have a huge-memory
machine to test these changes with, but will be happy to work with
folks that do and hps if this changes turns out not to be sufficient.

Submitted by: alfred@ from Hans Peter Selasky's original

15 years agoEliminate OBJ_NEEDGIANT. After r188331, OBJ_NEEDGIANT's only use is by a
alc [Sun, 8 Feb 2009 22:17:24 +0000 (22:17 +0000)]
Eliminate OBJ_NEEDGIANT.  After r188331, OBJ_NEEDGIANT's only use is by a
redundant assertion in vm_fault().

Reviewed by: kib

15 years agoRemove an overzealous check.
trasz [Sun, 8 Feb 2009 22:08:48 +0000 (22:08 +0000)]
Remove an overzealous check.

Submitted by: das
Reviewed by: scottl
Approved by: rwatson (mentor, implicit)
Sponsored by: FreeBSD Foundation

15 years agoFix multiple -s options.
kientzle [Sun, 8 Feb 2009 22:04:17 +0000 (22:04 +0000)]
Fix multiple -s options.

MFC after: 7 days

15 years agoWhen copying file data to the archive, don't write more
kientzle [Sun, 8 Feb 2009 22:02:46 +0000 (22:02 +0000)]
When copying file data to the archive, don't write more
than was read.  This seems to have only affected the shar
writer, since other formats proactively truncate output
to the originally-advertised size.

PR: bin/131244
MFC after: 7 days

15 years agoDon't try to directly update page tables
kmacy [Sun, 8 Feb 2009 21:54:51 +0000 (21:54 +0000)]
Don't try to directly update page tables

15 years agoRemove no longer valid comment.
kib [Sun, 8 Feb 2009 21:20:13 +0000 (21:20 +0000)]
Remove no longer valid comment.

Submitted by: alc

15 years agoThe NRL value in the queue head must be zero for interrupt transfers. This was
thompsa [Sun, 8 Feb 2009 21:08:00 +0000 (21:08 +0000)]
The NRL value in the queue head must be zero for interrupt transfers. This was
diagnosed using a USB analyser obtained through a FreeBSD Foundation grant.

Thanks to: FreeBSD Foundation

15 years agoImprove comments, correct English.
kib [Sun, 8 Feb 2009 20:52:09 +0000 (20:52 +0000)]
Improve comments, correct English.

Submitted by: alc

15 years agoDo not call vm_object_deallocate() from vm_map_delete(), because we
kib [Sun, 8 Feb 2009 20:39:17 +0000 (20:39 +0000)]
Do not call vm_object_deallocate() from vm_map_delete(), because we
hold the map lock there, and might need the vnode lock for OBJT_VNODE
objects. Postpone object deallocation until caller of vm_map_delete()
drops the map lock. Link the map entries to be freed into the freelist,
that is released by the new helper function vm_map_entry_free_freelist().

Reviewed by: tegge, alc
Tested by: pho

15 years agoIn vm_map_sync(), do not call vm_object_sync() while holding map lock.
kib [Sun, 8 Feb 2009 20:30:51 +0000 (20:30 +0000)]
In vm_map_sync(), do not call vm_object_sync() while holding map lock.
Reference object, drop the map lock, and then call vm_object_sync().
The object sync might require vnode lock for OBJT_VNODE type objects.

Reviewed by: tegge
Tested by: pho

15 years agofix typo
sam [Sun, 8 Feb 2009 20:29:37 +0000 (20:29 +0000)]
fix typo

Submitted by: Christoph Mallon <christoph.mallon@gmx.de>

15 years agoDo not sleep for vnode lock while holding map lock in vm_fault. Try to
kib [Sun, 8 Feb 2009 20:23:46 +0000 (20:23 +0000)]
Do not sleep for vnode lock while holding map lock in vm_fault. Try to
acquire vnode lock for OBJT_VNODE object after map lock is dropped.
Because we have the busy page(s) in the object, sleeping there would
result in deadlock with vnode resize. Try to get lock without sleeping,
and, if the attempt failed, drop the state, lock the vnode, and restart
the fault handler from the start with already locked vnode.

Because the vnode_pager_lock() function is inlined in vm_fault(),
axe it.

Based on suggestion by: alc
Reviewed by: tegge, alc
Tested by: pho

15 years agoPrefer the start and end attributes over the offset and size
marcel [Sun, 8 Feb 2009 20:19:19 +0000 (20:19 +0000)]
Prefer the start and end attributes over the offset and size
attributes. The start and end more accurately describe the
space taken by a partition. The offset and size are used to
describe the effective (usable) storage of that partition.

15 years agoo Add the "PART::scheme" attribute that returns the name of the
marcel [Sun, 8 Feb 2009 20:15:08 +0000 (20:15 +0000)]
o  Add the "PART::scheme" attribute that returns the name of the
   underlying partitioning scheme.
o  Put the start and end of the partition in the XML configuration.
   The start and end are the LBAs of the first and last sector
   (resp.) of the partition. They are currently identical to the
   offset and size attributes, which describe the partition as an
   offset and size in bytes, but may not in the future. The start
   and end will be used for the logical partition boundaries and
   may include metadata. The offset and size will always represent
   the useful storage space within the partition. Typically these
   two notions are the same, but for logical partitions in an
   extended partition, the EBR is more naturally treated as being
   part of the partition.

15 years agoo use humanize_number to print totals
sam [Sun, 8 Feb 2009 20:02:58 +0000 (20:02 +0000)]
o use humanize_number to print totals
o shrink some fields now that we don't have to worry about overflow

15 years agoAdd the comments to vm_map_simplify_entry() and vmspace_fork(),
kib [Sun, 8 Feb 2009 20:00:33 +0000 (20:00 +0000)]
Add the comments to vm_map_simplify_entry() and vmspace_fork(),
describing why several calls to vm_deallocate_object() with locked map
do not result in the acquisition of the vnode lock after map lock.

Suggested and reviewed by: tegge

15 years agoLock the new map in vmspace_fork(). The newly allocated map should not
kib [Sun, 8 Feb 2009 19:55:03 +0000 (19:55 +0000)]
Lock the new map in vmspace_fork(). The newly allocated map should not
be accessible outside vmspace_fork() yet, but locking it would satisfy
the protocol of the vm_map_entry_link() and other functions called
from vmspace_fork().

Use trylock that is supposedly cannot fail, to silence WITNESS warning
of the nested acquisition of the sx lock with the same name.

Suggested and reviewed by: tegge

15 years agoAssert that vnode is exclusively locked when its vm object is resized.
kib [Sun, 8 Feb 2009 19:44:50 +0000 (19:44 +0000)]
Assert that vnode is exclusively locked when its vm object is resized.

Reviewed by: tegge

15 years agoDo not leak the MAP_ENTRY_IN_TRANSITION flag when copying map entry
kib [Sun, 8 Feb 2009 19:41:08 +0000 (19:41 +0000)]
Do not leak the MAP_ENTRY_IN_TRANSITION flag when copying map entry
on fork. Otherwise, copied entry cannot be removed in the child map.

Reviewed by: tegge
MFC after: 2 weeks

15 years agoStyle.
kib [Sun, 8 Feb 2009 19:37:01 +0000 (19:37 +0000)]
Style.

15 years agoLookup up the directory entry for the tmpfs node that are deleted by
kib [Sun, 8 Feb 2009 19:18:33 +0000 (19:18 +0000)]
Lookup up the directory entry for the tmpfs node that are deleted by
both node pointer and name component. This does the right thing for
hardlinks to the same node in the same directory.

Submitted by: Yoshihiro Ota <ota j email ne jp>
PR: kern/131356
MFC after: 2 weeks

15 years agoShutup warning for DNAME RR.
ume [Sun, 8 Feb 2009 16:58:05 +0000 (16:58 +0000)]
Shutup warning for DNAME RR.

PR: bin/127591
Submitted by: "Eugene M. Kim" <20080111.freebsd.org__at__ab.ote.we.lv>
MFC after: 1 week

15 years agoSet the lower bound on queue size for an audit pipe to 1 instead of 0,
rwatson [Sun, 8 Feb 2009 15:38:31 +0000 (15:38 +0000)]
Set the lower bound on queue size for an audit pipe to 1 instead of 0,
as an audit pipe with a queue length of 0 is less useful.

Obtained from: TrustedBSD Project
Sponsored by: Apple, Inc.
MFC after: 1 week

15 years agoFix several filenames for "make clean" target.
wkoszek [Sun, 8 Feb 2009 14:43:20 +0000 (14:43 +0000)]
Fix several filenames for "make clean" target.

15 years agoChange various routines that are responsible for transforming audit
rwatson [Sun, 8 Feb 2009 14:39:35 +0000 (14:39 +0000)]
Change various routines that are responsible for transforming audit
event IDs based on arguments to return au_event_t rather than int.

Obtained from: TrustedBSD Project
Sponsored by: Apple, Inc.
MFC after: 1 week

15 years agoAudit AUE_MAC_EXECVE; currently just the standard AUE_EXECVE arguments
rwatson [Sun, 8 Feb 2009 14:24:35 +0000 (14:24 +0000)]
Audit AUE_MAC_EXECVE; currently just the standard AUE_EXECVE arguments
and not the label.

Obtained from: TrustedBSD Project
Sponsored by: Apple, Inc.
MFC after: 1 week

15 years agoAudit the flag argument to the nfssvc(2) system call.
rwatson [Sun, 8 Feb 2009 14:04:08 +0000 (14:04 +0000)]
Audit the flag argument to the nfssvc(2) system call.

Obtained from: TrustedBSD Project
Sponsored by: Apple, Inc.

15 years agosi(4) seems to build without a problem. However, since noone noticed lack
wkoszek [Sun, 8 Feb 2009 12:40:33 +0000 (12:40 +0000)]
si(4) seems to build without a problem. However, since noone noticed lack
of this driver, put it in a comment.

15 years agoFurther NOTES cleanup -- following drivers didn't survive TTY-ng
wkoszek [Sun, 8 Feb 2009 12:33:05 +0000 (12:33 +0000)]
Further NOTES cleanup -- following drivers didn't survive TTY-ng
and aren't included in NOTES anyway: cy(4), rc(4), rp(4).

si(4) doesn't belong to global NOTES.

15 years agoAdd missing pcfclock description.
wkoszek [Sun, 8 Feb 2009 12:12:19 +0000 (12:12 +0000)]
Add missing pcfclock description.