]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years ago- Remove a recently added comment from kernel_sysctlbyname() that isn't
John Baldwin [Tue, 10 Mar 2009 17:00:28 +0000 (17:00 +0000)]
- Remove a recently added comment from kernel_sysctlbyname() that isn't
  needed.
- Move the release of the sysctl sx lock after the vsunlock() in
  userland_sysctl() to restore the original memlock behavior of
  minimizing the amount of memory wired to handle sysctl requests.

MFC after: 1 week

15 years agocatch up with r189306; handle delayed activation of resources
Sam Leffler [Tue, 10 Mar 2009 16:42:49 +0000 (16:42 +0000)]
catch up with r189306; handle delayed activation of resources

Submitted by: jhb

15 years agoRemove these files, they refer to module bundles that do not exist anymore.
Andrew Thompson [Tue, 10 Mar 2009 15:54:37 +0000 (15:54 +0000)]
Remove these files, they refer to module bundles that do not exist anymore.

15 years agoUpdate libusb.3 name and add mlinks for usb.3 and libusb20.3
Andrew Thompson [Tue, 10 Mar 2009 15:49:43 +0000 (15:49 +0000)]
Update libusb.3 name and add mlinks for usb.3 and libusb20.3

15 years agoAdd an ABI compat shim for the vfs.bufspace sysctl for sysctl requests that
John Baldwin [Tue, 10 Mar 2009 15:26:50 +0000 (15:26 +0000)]
Add an ABI compat shim for the vfs.bufspace sysctl for sysctl requests that
try to fetch it as an int rather than a long.  If the current value is
greater than INT_MAX it reports a value of INT_MAX.

15 years agoUpdate top and systat for vfs.bufcache now being a long rather than an int.
John Baldwin [Tue, 10 Mar 2009 15:25:19 +0000 (15:25 +0000)]
Update top and systat for vfs.bufcache now being a long rather than an int.

15 years agoWhen attaching a geli on boot make sure that it is detached
Guido van Rooij [Tue, 10 Mar 2009 15:23:43 +0000 (15:23 +0000)]
When attaching a geli on boot make sure that it is detached
upon last close. (needed for a gmirror to properly shutdown
upon reboot when a geli is on top the gmirror)

15 years agoWhen swap resides on a mirror and it is not stopped, the mirror
Guido van Rooij [Tue, 10 Mar 2009 15:19:49 +0000 (15:19 +0000)]
When swap resides on a mirror and it is not stopped, the mirror
is degraded upon the next reboot and will have to be rebuild.
Thus call swapoff when rebooting (read: when stopping swap1)

15 years agoAdd tcpp -- TCP parallelism microbenchmark.
Robert Watson [Tue, 10 Mar 2009 14:52:17 +0000 (14:52 +0000)]
Add tcpp -- TCP parallelism microbenchmark.

This tool creates large numbers of TCP connections, each of which will
transmit a fixed amount of data, between client and server hosts.  tcpp can
use multiple workers (typically up to the number of hardware cores), and can
use multiple source IPs in order to use an expanded port/IP 4-tuple space to
avoid problems from reusing 4-tuples too quickly.  Aggregate bandwidth use
will be reported after a client run.

While by no means a perfect tool, it has proven quite useful in generating
and optimizing TCP stack lock contention by easily generating high-intensity
workloads.  It also proves surprisingly good at finding device driver bugs.

15 years agoDo not use bypass for vop_vptocnp() from nullfs, call standard
Konstantin Belousov [Tue, 10 Mar 2009 14:35:21 +0000 (14:35 +0000)]
Do not use bypass for vop_vptocnp() from nullfs, call standard
implementation instead. The bypass does not assume that returned vnode
is only held.

Reported by: Paul B. Mahol <onemda gmail com>, pluknet <pluknet gmail com>
Reviewed by: jhb
Tested by: pho, pluknet <pluknet gmail com>

15 years agoBe compatible with LibUSB from sourceforge and close the handle after reset
Andrew Thompson [Tue, 10 Mar 2009 14:29:34 +0000 (14:29 +0000)]
Be compatible with LibUSB from sourceforge and close the handle after reset

Submitted by: Hans Petter Selasky

15 years agoDisable zerocopy by default for now. It's causing some problems in pcap
Christian S.J. Peron [Tue, 10 Mar 2009 14:28:19 +0000 (14:28 +0000)]
Disable zerocopy by default for now.  It's causing some problems in pcap
consumers which fork after the shared pages have been setup.  pflogd(8)
is an example.  The problem is understood and there is a fix coming in
shortly.

Folks who want to continue using it can do so by setting

net.bpf.zerocopy_enable

to 1.

Discussed with: rwatson

15 years agoWhen freeing all the resources of the card, it is better to turn off
Warner Losh [Tue, 10 Mar 2009 12:10:50 +0000 (12:10 +0000)]
When freeing all the resources of the card, it is better to turn off
the PORTEN and MEMEN bits in the command register than to zero the
bars.

Use pci_write_ivar directly instead of a one-line wrapper that adds no
value.

Track verbosity changes in pci.

Remove a stray blank line.

15 years agoMerge r183430 from vendor/top/dist to head/contrib/top, although with
Robert Watson [Tue, 10 Mar 2009 11:46:41 +0000 (11:46 +0000)]
Merge r183430 from vendor/top/dist to head/contrib/top, although with
record-only mergeinfo because an automated merge is confused by the
flattening that took place:

  Move install to install-sh to prevent name-clashes.

MFC after: 3 days

15 years agoMake a 1:1 mapping between syscons stats and terminal emulators.
Ed Schouten [Tue, 10 Mar 2009 11:28:54 +0000 (11:28 +0000)]
Make a 1:1 mapping between syscons stats and terminal emulators.

After I imported libteken into the source tree, I noticed syscons didn't
store the cursor position inside the terminal emulator, but inside the
virtual terminal stat. This is not very useful, because when you
implement more complex forms of line wrapping, you need to keep track of
more state than just the cursor position.

Because the kernel messages didn't share the same terminal emulator as
ttyv0, this caused a lot of strange things, like kernel messages being
misplaced and a missing notification to resize the terminal emulator for
kernel messages never to be resized when using vidcontrol.

This patch just removes kernel_console_ts and adds a special parameter
to te_puts to determine whether messages should be printed using regular
colors or the ones for kernel messages.

Reported by: ache
Tested by: nyan, garga (older version)

15 years agoRestore the return statement. It was accidentally removed by rev 188429.
Yoshihiro Takahashi [Tue, 10 Mar 2009 11:14:03 +0000 (11:14 +0000)]
Restore the return statement.  It was accidentally removed by rev 188429.

15 years agoRemove now-unused INP_UNMAPPABLEOPTS.
Robert Watson [Tue, 10 Mar 2009 11:04:19 +0000 (11:04 +0000)]
Remove now-unused INP_UNMAPPABLEOPTS.

MFC after: 3 days
Discussed with: bz

15 years agoRename files that collide on case-insensitive file systems by encoding
Robert Watson [Tue, 10 Mar 2009 09:33:22 +0000 (09:33 +0000)]
Rename files that collide on case-insensitive file systems by encoding
colliding upper case letters as the lower case letter with a '_' in
front.

MFC after: 3 days
Discussed with: ed
Spotted by: Michael David Crawford <mdc at prgmr.com>

15 years agoFix a buglet in revision 189401: when restoring a 64-bit BAR,
Marcel Moolenaar [Tue, 10 Mar 2009 06:21:52 +0000 (06:21 +0000)]
Fix a buglet in revision 189401: when restoring a 64-bit BAR,
write the upper 32-bits in the adjacent bar. The consequences
of the buglet were severe enough though: a machine check.

15 years agoEliminate the last use of the recursive mapping to access user-space page
Alan Cox [Tue, 10 Mar 2009 02:12:03 +0000 (02:12 +0000)]
Eliminate the last use of the recursive mapping to access user-space page
table pages.  Now, all accesses to user-space page table pages are
performed through the direct map.  (The recursive mapping is only used
to access kernel-space page table pages.)

Eliminate the TLB invalidation on the recursive mapping when a user-space
page table page is removed from the page table and when a user-space
superpage is demoted.

15 years agoadd cfid and geom_redboot
Sam Leffler [Mon, 9 Mar 2009 23:25:34 +0000 (23:25 +0000)]
add cfid and geom_redboot

15 years agoadd geom_redboot, a geom module that exports RedBoot FIS partitions as named
Sam Leffler [Mon, 9 Mar 2009 23:18:36 +0000 (23:18 +0000)]
add geom_redboot, a geom module that exports RedBoot FIS partitions as named
slices in dev/redboot/*

15 years agoFix up the entries for libusb, it seems it existed back in 2002 so it was
Andrew Thompson [Mon, 9 Mar 2009 23:18:07 +0000 (23:18 +0000)]
Fix up the entries for libusb, it seems it existed back in 2002 so it was
getting removed again.

Reported by: Steve Kargl

15 years agoAdd cfid, a disk interface to CFI flash devices; this enables construction
Sam Leffler [Mon, 9 Mar 2009 23:16:02 +0000 (23:16 +0000)]
Add cfid, a disk interface to CFI flash devices; this enables construction
of flash-based filesystems.

Note this is not interlocked against the raw CFI device.

15 years agoreplace if_watchdog w/ private callout; probably can merge this with the
Sam Leffler [Mon, 9 Mar 2009 23:10:19 +0000 (23:10 +0000)]
replace if_watchdog w/ private callout; probably can merge this with the
calibration work sometime in the future

15 years agoremove ar9160Detach; it does nothing
Sam Leffler [Mon, 9 Mar 2009 23:04:06 +0000 (23:04 +0000)]
remove ar9160Detach; it does nothing

15 years agoFix uninitialized use of ifp for ii.
Bruce M Simpson [Mon, 9 Mar 2009 22:54:17 +0000 (22:54 +0000)]
Fix uninitialized use of ifp for ii.

Found by: Peter Holm

15 years agoFix spelling.
Andrew Thompson [Mon, 9 Mar 2009 22:43:00 +0000 (22:43 +0000)]
Fix spelling.

15 years agoUpdate 20090309 to say that libmap.conf entries for libusb are no longer needed.
Andrew Thompson [Mon, 9 Mar 2009 22:42:01 +0000 (22:42 +0000)]
Update 20090309 to say that libmap.conf entries for libusb are no longer needed.

15 years agoAdd type specific suspend/resume ata channel functions. Add checks to avoid
Alexander Motin [Mon, 9 Mar 2009 20:48:57 +0000 (20:48 +0000)]
Add type specific suspend/resume ata channel functions. Add checks to avoid
crash on detached channel resume. Add placeholder for possible type-specific
suspend/resume routines.

15 years agoMFp4 //depot/projects/usb 158942,158948
Andrew Thompson [Mon, 9 Mar 2009 20:08:08 +0000 (20:08 +0000)]
MFp4 //depot/projects/usb 158942,158948

Allow USB to be compiled without ugen support.

Submitted by: Hans Petter Selasky

15 years agoFix musb_otg.h include filename.
Andrew Thompson [Mon, 9 Mar 2009 20:05:46 +0000 (20:05 +0000)]
Fix musb_otg.h include filename.

Submitted by: Hans Petter Selasky

15 years ago- Fix a typo.
Stanislav Sedov [Mon, 9 Mar 2009 19:56:37 +0000 (19:56 +0000)]
- Fix a typo.

Spotted by: Steve Kargl <sgk@troutmask.apl.washington.edu>

15 years agoDon't call into the TTY layer when inside kdb.
Ed Schouten [Mon, 9 Mar 2009 19:46:19 +0000 (19:46 +0000)]
Don't call into the TTY layer when inside kdb.

We should just leave the underlying TTY objects alone when scrolling
around in KDB. It should be handled by Syscons exclusively.

Reported by: pluknet gmail com

15 years agoAdjust some variables (mostly related to the buffer cache) that hold
John Baldwin [Mon, 9 Mar 2009 19:35:20 +0000 (19:35 +0000)]
Adjust some variables (mostly related to the buffer cache) that hold
address space sizes to be longs instead of ints.  Specifically, the follow
values are now longs: runningbufspace, bufspace, maxbufspace,
bufmallocspace, maxbufmallocspace, lobufspace, hibufspace, lorunningspace,
hirunningspace, maxswzone, maxbcache, and maxpipekva.  Previously, a
relatively small number (~ 44000) of buffers set in kern.nbuf would result
in integer overflows resulting either in hangs or bogus values of
hidirtybuffers and lodirtybuffers.  Now one has to overflow a long to see
such problems.  There was a check for a nbuf setting that would cause
overflows in the auto-tuning of nbuf.  I've changed it to always check and
cap nbuf but warn if a user-supplied tunable would cause overflow.

Note that this changes the ABI of several sysctls that are used by things
like top(1), etc., so any MFC would probably require a some gross shims
to allow for that.

MFC after: 1 month

15 years ago- Point libusb users to the ports collection UPDATING file.
Stanislav Sedov [Mon, 9 Mar 2009 19:22:45 +0000 (19:22 +0000)]
- Point libusb users to the ports collection UPDATING file.

15 years agoMove the debug.hashstat sysctl tree under DIAGNOSTIC. I measured the
John Baldwin [Mon, 9 Mar 2009 19:04:53 +0000 (19:04 +0000)]
Move the debug.hashstat sysctl tree under DIAGNOSTIC.  I measured the
debug.hashstat.rawnchash sysctl in particular as taking 7 milliseconds on
a 3GHz Intel Xeon (4x2) running 7.1.  It accounted for almost a quarter of
the total runtime of 'sysctl -a'.  It also performs lots of copyout's while
holding the namecache lock (this does not attempt to fix that).

MFC after: 2 weeks

15 years agoMerge IGMPv3 and Source-Specific Multicast (SSM) to the FreeBSD
Bruce M Simpson [Mon, 9 Mar 2009 17:53:05 +0000 (17:53 +0000)]
Merge IGMPv3 and Source-Specific Multicast (SSM) to the FreeBSD
IPv4 stack.

Diffs are minimized against p4.
PCS has been used for some protocol verification, more widespread
testing of recorded sources in Group-and-Source queries is needed.
sizeof(struct igmpstat) has changed.

__FreeBSD_version is bumped to 800070.

15 years agoMark the bsdextended rules sysctl as being mpsafe.
Christian S.J. Peron [Mon, 9 Mar 2009 17:42:18 +0000 (17:42 +0000)]
Mark the bsdextended rules sysctl as being mpsafe.

Discussed with: rwatson

15 years agoCommit missed file in r189587, update directory name for libusb.
Andrew Thompson [Mon, 9 Mar 2009 17:38:14 +0000 (17:38 +0000)]
Commit missed file in r189587, update directory name for libusb.

Spotted by: rdivacky

15 years ago- Make it possible to disable GPT support by setting LOADER_NO_GPT_SUPPORT
John Baldwin [Mon, 9 Mar 2009 17:16:29 +0000 (17:16 +0000)]
- Make it possible to disable GPT support by setting LOADER_NO_GPT_SUPPORT
  in make.conf or src.conf.
- When GPT is enabled (which it is by default), use memory above 1 MB and
  leave the memory from the end of the bss to the end of the 640k window
  purely for the stack.  The loader has grown and now it is much more
  common for the heap and stack to grow into each other when both are
  located in the 640k window.

PR: kern/129526
MFC after: 1 week

15 years agolibusb20 is now installed as libusb, remove the version number from the
Andrew Thompson [Mon, 9 Mar 2009 17:09:46 +0000 (17:09 +0000)]
libusb20 is now installed as libusb, remove the version number from the
directory name.

15 years agoInstall libusb20.so.1 as libusb.so.1, there will be a followup commit to the
Andrew Thompson [Mon, 9 Mar 2009 17:05:31 +0000 (17:05 +0000)]
Install libusb20.so.1 as libusb.so.1, there will be a followup commit to the
ports tree so that programs use libusb from the base by default. Thanks to
Stanislav Sedov for sorting out the ports build.

Bump __FreeBSD_version to 800069

Help and testing by: stas

15 years agoAdd igmp(4) man page, do not connect to build yet.
Bruce M Simpson [Mon, 9 Mar 2009 16:51:40 +0000 (16:51 +0000)]
Add igmp(4) man page, do not connect to build yet.

15 years agoMFp4 //depot/projects/usb@158916
Andrew Thompson [Mon, 9 Mar 2009 15:25:46 +0000 (15:25 +0000)]
MFp4 //depot/projects/usb@158916

USB mouse patch to address complicated data reporting descriptors.

Reported by: Boris Kotzev
Submitted by: Hans Petter Selasky

15 years agoPrefer prototypes to k&r definitions.
Warner Losh [Mon, 9 Mar 2009 13:32:19 +0000 (13:32 +0000)]
Prefer prototypes to k&r definitions.

15 years agoMake generic_intr routines match prototype.
Warner Losh [Mon, 9 Mar 2009 13:30:00 +0000 (13:30 +0000)]
Make generic_intr routines match prototype.

15 years ago__LP64__ is what's defined, not _LP64_, according to the manual (and
Warner Losh [Mon, 9 Mar 2009 13:29:13 +0000 (13:29 +0000)]
__LP64__ is what's defined, not _LP64_, according to the manual (and
also experience).

15 years agoFix prototypes to be consistent.
Warner Losh [Mon, 9 Mar 2009 13:27:33 +0000 (13:27 +0000)]
Fix prototypes to be consistent.

15 years agoo Add declarations for a few more nodes widely used.
Warner Losh [Mon, 9 Mar 2009 13:26:55 +0000 (13:26 +0000)]
o Add declarations for a few more nodes widely used.
o Minor formatting nit.

15 years agoremove now-redunant cardbus attachment lines.
Warner Losh [Mon, 9 Mar 2009 13:25:34 +0000 (13:25 +0000)]
remove now-redunant cardbus attachment lines.

15 years agoremove now-redunant cardbus attachment.
Warner Losh [Mon, 9 Mar 2009 13:23:54 +0000 (13:23 +0000)]
remove now-redunant cardbus attachment.

15 years agoFix a long-standing bug in newbus. It was introduced when subclassing
Warner Losh [Mon, 9 Mar 2009 13:20:23 +0000 (13:20 +0000)]
Fix a long-standing bug in newbus.  It was introduced when subclassing
was introduced.  If you have a bus, say cardbus, that is derived from
a base-bus (say PCI), then ordinarily all PCI drivers would attach to
cardbus devices.  However, there had been one exception: kldload
wouldn't work.

The problem is in devclass_add_driver.  In this routine, all we did
was call to the pci device's BUS_DRIVER_ADDED routine.  However, since
cardbus bus instances had a different devclass, none of them were
called.

The solution is to call all subclass devclasses, recursively down the
tree, of the class that was loaded.  Since we don't have a 'children
class' pointer, we search the whole list of devclasses for a class
whose parent matches.  Since just done a kldload time, this isn't as
bad as it sounds.  In addition, we short-circuit the whole process by
marking those classes with subclasses with a flag.  We'll likely have
to reevaluate this method the number of devclasses with subclasses
gets large.

This means we can remove the "cardbus" lines from all the PCI drivers
since we have no cardbus specific attach device attachments in the
tree.

# Also: minor tweak to an error message

15 years agoUse a u_int for p_lock instead of a char: this avoids a (somewhat
Robert Watson [Mon, 9 Mar 2009 13:12:48 +0000 (13:12 +0000)]
Use a u_int for p_lock instead of a char: this avoids a (somewhat
unlikely but not impossible given modern thread counts) wrap-around,
and the compiler was padding it out to an int (at least) anyway.

MFC after: 3 days (but confirm ABI impact)

15 years agoTrim comments about the MP-safety of various bits of the amd64/i386
Robert Watson [Mon, 9 Mar 2009 13:11:16 +0000 (13:11 +0000)]
Trim comments about the MP-safety of various bits of the amd64/i386
system call entry path and i386 IP checksum generation: we now assume
all code is MPSAFE unless explicitly marked otherwise.  Remove XXX
Giant comments along similar lines: the code by the comments either
doesn't need or doesn't want Giant (especially the NMI handler).

MFC after: 3 days

15 years agoRemove two now-defunct KSE fields from struct thread: td_uuticks and
Robert Watson [Mon, 9 Mar 2009 11:18:41 +0000 (11:18 +0000)]
Remove two now-defunct KSE fields from struct thread: td_uuticks and
td_usticks.

15 years agoAdd a new thread-private flag, TDP_AUDITREC, to indicate whether or
Robert Watson [Mon, 9 Mar 2009 10:45:58 +0000 (10:45 +0000)]
Add a new thread-private flag, TDP_AUDITREC, to indicate whether or
not there is an audit record hung off of td_ar on the current thread.
Test this flag instead of td_ar when auditing syscall arguments or
checking for an audit record to commit on syscall return.  Under
these circumstances, td_pflags is much more likely to be in the cache
(especially if there is no auditing of the current system call), so
this should help reduce cache misses in the system call return path.

MFC after:      1 week
Reported by:    kris
Obtained from:  TrustedBSD Project

15 years agoFor IP1001 PHYs, read auto-negotiation advertisement register to
Pyun YongHyeon [Mon, 9 Mar 2009 08:17:46 +0000 (08:17 +0000)]
For IP1001 PHYs, read auto-negotiation advertisement register to
get default next page configuration. While I'm here explicitly set
IP1000PHY_ANAR_CSMA bit. This bit is read-only and always set
by hardware so setting it has no effect but it would clear the
intention. With this change controllers that couldn't establish
1000baseT link should work.

PR: kern/130846

15 years agoUse mii_phy_add_media() and remove setting each media type.
Pyun YongHyeon [Mon, 9 Mar 2009 08:09:06 +0000 (08:09 +0000)]
Use mii_phy_add_media() and remove setting each media type.
While I'm here, don't set mii_anegticks as it's set by
mii_phy_add_media().

15 years agoFor unknown speed, explicitly set IFM_NONE.
Pyun YongHyeon [Mon, 9 Mar 2009 08:01:40 +0000 (08:01 +0000)]
For unknown speed, explicitly set IFM_NONE.

15 years agoReport current link state while auto-negotiation is in progress.
Pyun YongHyeon [Mon, 9 Mar 2009 07:56:40 +0000 (07:56 +0000)]
Report current link state while auto-negotiation is in progress.

15 years agoConsistently use kdev for the kernel device.
Robert Noland [Mon, 9 Mar 2009 07:55:18 +0000 (07:55 +0000)]
Consistently use kdev for the kernel device.

Submitted by: vehemens <vehemens@verizon.net>
MFC after: 3 days

15 years agoClean up the printing on amd64. Should also be consistent on i386.
Robert Noland [Mon, 9 Mar 2009 07:50:27 +0000 (07:50 +0000)]
Clean up the printing on amd64.  Should also be consistent on i386.

MFC after: 3 days

15 years agoThere is no need to sync these buffers to swap.
Robert Noland [Mon, 9 Mar 2009 07:49:13 +0000 (07:49 +0000)]
There is no need to sync these buffers to swap.

MFC after: 3 days

15 years agoChange the flags to bus_dmamem around to allow it to sleep waiting for
Robert Noland [Mon, 9 Mar 2009 07:47:03 +0000 (07:47 +0000)]
Change the flags to bus_dmamem around to allow it to sleep waiting for
resources during allocation, but not during map load.  Also, zero the
buffers here.

MFC after: 3 days

15 years agoFix the flags to bus_dmamem_* to allow the allocation to sleep while
Robert Noland [Mon, 9 Mar 2009 07:38:22 +0000 (07:38 +0000)]
Fix the flags to bus_dmamem_* to allow the allocation to sleep while
waiting for resources.  It is really the load that we can't defer.
BUS_DMA_NOCACHE belongs on bus_dmamap_load() as well.

MFC after: 3 days

15 years ago -Make the PCI(E)/AGP calculations consistent
Robert Noland [Mon, 9 Mar 2009 07:33:35 +0000 (07:33 +0000)]
-Make the PCI(E)/AGP calculations consistent
-Calculate the scratch address correctly

MFC after: 10 days

15 years agoCall the right function for the right chipset.
Robert Noland [Mon, 9 Mar 2009 07:24:32 +0000 (07:24 +0000)]
Call the right function for the right chipset.

MFC after: 10 days

15 years agoAdd a new tunable hw.re.prefer_iomap which disables memory register
Pyun YongHyeon [Mon, 9 Mar 2009 06:02:55 +0000 (06:02 +0000)]
Add a new tunable hw.re.prefer_iomap which disables memory register
mapping. The tunable is OFF for all controllers except RTL8169SC
family. RTL8169SC seems to require more magic to use memory
register mapping. r187483 added a fix for RTL8169SCe controller but
it does not looke like fix other variants of RTL8169SC.

Tested by: Gavin Stone-Tolcher g.stone-tolcher <> its dot uq dot edu dot au

15 years agoDon't ignore other fcntl functions, directly call __sys_fcntl if
David Xu [Mon, 9 Mar 2009 05:54:43 +0000 (05:54 +0000)]
Don't ignore other fcntl functions, directly call __sys_fcntl if
WITHOUT_SYSCALL_COMPAT is not defined.

Reviewed by: deischen

15 years agoA system with plenty of memory would not require so much swap for generic
Xin LI [Mon, 9 Mar 2009 05:41:04 +0000 (05:41 +0000)]
A system with plenty of memory would not require so much swap for generic
usage.

Discussed with: dillon

15 years agoChange pmap_enter_quick_locked() so that it uses the kernel's direct map
Alan Cox [Mon, 9 Mar 2009 03:35:25 +0000 (03:35 +0000)]
Change pmap_enter_quick_locked() so that it uses the kernel's direct map
instead of the pmap's recursive mapping to access the lowest level of the
page table when it maps a user-space virtual address.

15 years agoFix TXPMGT handling:
Sam Leffler [Mon, 9 Mar 2009 02:37:52 +0000 (02:37 +0000)]
Fix TXPMGT handling:
o correct dBm<->mW conversion logic
o set net80211 TXPMGT capability only if driver reports it is capable

PR: kern/132342
Submitted by: "Paul B. Mahol" <onemda@gmail.com>

15 years agoDon't reference non-existent __fcntl_compat if WITHOUT_SYSCALL_COMPAT is defined.
David Xu [Mon, 9 Mar 2009 02:34:02 +0000 (02:34 +0000)]
Don't reference non-existent __fcntl_compat if WITHOUT_SYSCALL_COMPAT is defined.

Submitted by: Pawel Worach "pawel dot worach at gmail dot com"

15 years agoo mark unexpected callbacks more clearly
Sam Leffler [Sun, 8 Mar 2009 23:45:56 +0000 (23:45 +0000)]
o mark unexpected callbacks more clearly
o unwrap some lines

15 years agoMFp4 //depot/projects/usb@158868
Andrew Thompson [Sun, 8 Mar 2009 22:58:19 +0000 (22:58 +0000)]
MFp4 //depot/projects/usb@158868

Fix bugs and improve HID parsing.
- fix possible memory leak found
- fix possible NULL pointer access
- fix possible invalid memory read
- parsing improvements
- reset item data position when a new report ID is detected.

Submitted by: Hans Petter Selasky

15 years agoMFp4 //depot/projects/usb@158869
Andrew Thompson [Sun, 8 Mar 2009 22:55:17 +0000 (22:55 +0000)]
MFp4 //depot/projects/usb@158869

Fix sael init code.

Reported by: Alberto Mijares
Submitted by: Hans Petter Selasky

15 years agoBy default, don't compile in counters of calls to various time
Robert Watson [Sun, 8 Mar 2009 22:19:28 +0000 (22:19 +0000)]
By default, don't compile in counters of calls to various time
query functions in the kernel, as these effectively serialize
parallel calls to the gettimeofday(2) system call, as well as
other kernel services that use timestamps.

Use the NetBSD version of the fix (kern_tc.c:1.32 by ad@) as
they have picked up our timecounter code and also ran into the
same problem.

Reported by: kris
Obtained from: NetBSD
MFC after: 3 days

15 years agoDecompose the global UNIX domain sockets rwlock into two different
Robert Watson [Sun, 8 Mar 2009 21:48:29 +0000 (21:48 +0000)]
Decompose the global UNIX domain sockets rwlock into two different
locks: a global list/counter/generation counter protected by a new
mutex unp_list_lock, and a global linkage rwlock, unp_global_rwlock,
which protects the connections between UNIX domain sockets.

This eliminates conditional lock acquisition that was previously a
property of the global lock being held over sonewconn() leading to a
call to uipc_attach(), which also required the global lock, but
couldn't rely on it as other paths existed to uipc_attach() that
didn't hold it: now uipc_attach() uses only the list lock, which
follows the linkage lock in the lock order.  It may also reduce
contention on the global lock for some workloads.

Add global UNIX domain socket locks to hard-coded witness lock
order.

MFC after: 1 week
Discussed with: kris

15 years agoIn UNIX domain socket GC regression test, after setting a socket
Robert Watson [Sun, 8 Mar 2009 21:06:02 +0000 (21:06 +0000)]
In UNIX domain socket GC regression test, after setting a socket
non-blocking, EINPROGRESS is an acceptable result from connect().

15 years agoDon't disable CR-to-NL translation when waiting for data to arrive.
Ed Schouten [Sun, 8 Mar 2009 19:09:55 +0000 (19:09 +0000)]
Don't disable CR-to-NL translation when waiting for data to arrive.

A difference between the old and the new TTY layer is that the new
implementation does not perform any post-processing before returning
data back to userspace when calling read().

sh(1)'s read turns the TTY into a raw mode before calling select(). This
means that the first character will not receive any ICRNL processing.
Inherit this flag from the original terminal attributes.

Even though this issue is not present on RELENG_*, I'm MFCing it to make
sh(1) in jails behave better.

PR: bin/129566
MFC after: 2 weeks

15 years agoDocument the new default implementation of VOP_VPTOCNP(9).
Joe Marcus Clarke [Sun, 8 Mar 2009 19:07:44 +0000 (19:07 +0000)]
Document the new default implementation of VOP_VPTOCNP(9).

Approved by: kib

15 years agoAdd a prototype for the new vop_stdvptocnp function.
Joe Marcus Clarke [Sun, 8 Mar 2009 19:06:26 +0000 (19:06 +0000)]
Add a prototype for the new vop_stdvptocnp function.

Reviewed by: kib
Approved by: kib
Tested by: pho

15 years agoAdd a default implementation for VOP_VPTOCNP(9) which scans the parent
Joe Marcus Clarke [Sun, 8 Mar 2009 19:05:53 +0000 (19:05 +0000)]
Add a default implementation for VOP_VPTOCNP(9) which scans the parent
directory of a vnode to find a dirent with a matching file number.  The
name from that dirent is then used to provide the component name.

Note: if the initial vnode argument is not a directory itself, then
the default VOP_VPTOCNP(9) implementation still returns ENOENT.

Reviewed by: kib
Approved by: kib
Tested by: pho

15 years agoo Spell. Sort .Xrs.
Maxim Konovalov [Sun, 8 Mar 2009 18:02:30 +0000 (18:02 +0000)]
o Spell.  Sort .Xrs.

15 years agoDocument several notifications, among them are DEVFS, update to ifneti,
Konstantin Belousov [Sun, 8 Mar 2009 14:28:40 +0000 (14:28 +0000)]
Document several notifications, among them are DEVFS, update to ifneti,
coretemp and kern.

The asmc(4) and zfs(5) are still not documented.

Based on the patch by Roland Smith <rsmith xs4all nl>.
MFC after: 1 week

15 years agoRemove 'uio' argument from MAC Framework and MAC policy entry points for
Robert Watson [Sun, 8 Mar 2009 12:32:06 +0000 (12:32 +0000)]
Remove 'uio' argument from MAC Framework and MAC policy entry points for
extended attribute get/set; in the case of get an uninitialized user
buffer was passed before the EA was retrieved, making it of relatively
little use; the latter was simply unused by any policies.

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

15 years agoRename 'ucred' argument to mac_socket_check_bind() to 'cred' to match
Robert Watson [Sun, 8 Mar 2009 12:22:00 +0000 (12:22 +0000)]
Rename 'ucred' argument to mac_socket_check_bind() to 'cred' to match
other use of the same variable type.

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

15 years agoImprove the consistency of MAC Framework and MAC policy entry point
Robert Watson [Sun, 8 Mar 2009 10:58:37 +0000 (10:58 +0000)]
Improve the consistency of MAC Framework and MAC policy entry point
naming by renaming certain "proc" entry points to "cred" entry points,
reflecting their manipulation of credentials.  For some entry points,
the process was passed into the framework but not into policies; in
these cases, stop passing in the process since we don't need it.

  mac_proc_check_setaudit -> mac_cred_check_setaudit
  mac_proc_check_setaudit_addr -> mac_cred_check_setaudit_addr
  mac_proc_check_setauid -> mac_cred_check_setauid
  mac_proc_check_setegid -> mac_cred_check_setegid
  mac_proc_check_seteuid -> mac_cred_check_seteuid
  mac_proc_check_setgid -> mac_cred_check_setgid
  mac_proc_check_setgroups -> mac_cred_ceck_setgroups
  mac_proc_check_setregid -> mac_cred_check_setregid
  mac_proc_check_setresgid -> mac_cred_check_setresgid
  mac_proc_check_setresuid -> mac_cred_check_setresuid
  mac_proc_check_setreuid -> mac_cred_check_setreuid
  mac_proc_check_setuid -> mac_cred_check_setuid

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

15 years agoMove m_getcl() into its own function. This also fixes a bug where the m_adj for
Andrew Thompson [Sun, 8 Mar 2009 06:56:13 +0000 (06:56 +0000)]
Move m_getcl() into its own function. This also fixes a bug where the m_adj for
ETHER_ALIGN was having no effect since m_len had not been set.

15 years agoUpdate version to 2.6.901a to indicate this is synced up with r745 of
Tim Kientzle [Sun, 8 Mar 2009 06:20:35 +0000 (06:20 +0000)]
Update version to 2.6.901a to indicate this is synced up with r745 of
libarchive.googlecode.com (except for the lzma/xz support).

15 years agoMerge r492 from libarchive.googlecode.com: First cut at exposing the
Tim Kientzle [Sun, 8 Mar 2009 06:19:28 +0000 (06:19 +0000)]
Merge r492 from libarchive.googlecode.com:  First cut at exposing the
new options mechanism to userland.  Documentation pending...

15 years agoDon't declare bin_search() as an inline function, since there's no
David Schultz [Sun, 8 Mar 2009 06:14:33 +0000 (06:14 +0000)]
Don't declare bin_search() as an inline function, since there's no
inline definition of it.

15 years agoMatch a comment to reduce differences with libarchive.googlecode.com.
Tim Kientzle [Sun, 8 Mar 2009 06:09:20 +0000 (06:09 +0000)]
Match a comment to reduce differences with libarchive.googlecode.com.

15 years agoMerge r709,r710 from libarchive.googlecode.com: More work on
Tim Kientzle [Sun, 8 Mar 2009 06:07:35 +0000 (06:07 +0000)]
Merge r709,r710 from libarchive.googlecode.com:  More work on
Windows support.

15 years agoFix endian conversion from htole16 to htole32.
Andrew Thompson [Sun, 8 Mar 2009 06:03:28 +0000 (06:03 +0000)]
Fix endian conversion from htole16 to htole32.

Tested with: ARM xscale

15 years agoMerge r687-689,691,693-701,720 from libarchive.googlecode.com:
Tim Kientzle [Sun, 8 Mar 2009 06:03:15 +0000 (06:03 +0000)]
Merge r687-689,691,693-701,720 from libarchive.googlecode.com:
Translate getdate.y into C for portability.  Make the get_date()
function easier to test as well:
 * Have it accept a time_t "now" to use as a reference so that test
   code can verify relative time specifications against known starting
   points.
 * Set up default date after parsing the string so that we
   can use the specified timezone (if any) instead of the local
   default.  Otherwise, local DST makes it almost impossible to
   reliably test time specifications such as "sunday UTC"

15 years agoMerger r629-631,633-646,648,654,678,681,682 from libarchive.googlecode.com:
Tim Kientzle [Sun, 8 Mar 2009 05:47:21 +0000 (05:47 +0000)]
Merger r629-631,633-646,648,654,678,681,682 from libarchive.googlecode.com:
Many changes for Windows compatibility.  bsdtar_test now runs successfully
on both POSIX platforms and Windows.

15 years agoMerge r368,496,625,626 from libarchive.googlecode.com: A number of
Tim Kientzle [Sun, 8 Mar 2009 05:38:45 +0000 (05:38 +0000)]
Merge r368,496,625,626 from libarchive.googlecode.com: A number of
style and portability tweaks to the test harness.  Most significantly,
don't use getopt().

15 years agoMerge r622 from libarchive.googlecode.com: Avoid warning on platforms
Tim Kientzle [Sun, 8 Mar 2009 05:35:59 +0000 (05:35 +0000)]
Merge r622 from libarchive.googlecode.com:  Avoid warning on platforms
that lack regex.h.