François Tigeot [Sat, 7 Feb 2015 21:41:45 +0000 (22:41 +0100)]
idr: Remove a debugging message
Sascha Wildner [Sun, 8 Feb 2015 00:15:26 +0000 (01:15 +0100)]
kernel/x86_64: Add SDBG to Features2.
From Intel 64 and IA-32 Architectures Software Developer's Manuals:
"A value of 1 indicates the processor supports IA32_DEBUG_INTERFACE MSR
for silicon debug."
Found in the Xeon E3-1226 v3.
François Tigeot [Sat, 7 Feb 2015 16:57:28 +0000 (17:57 +0100)]
drm: Reintroduce drm_kms_free()
* The Linux version of kfree() is a macro with one argument.
* The real kfree() function takes two arguments.
* kfree() cannot be simply called via a function pointer from
Linux code.
* Re-add the special purpose drm_kms_free() wrapper function
to handle this situation. It was wrongly removed when drm_crtc.c
was last synced to Linux.
Reported-by: jh32
François Tigeot [Sat, 7 Feb 2015 10:11:17 +0000 (11:11 +0100)]
drm: Start using idr_alloc()
François Tigeot [Sat, 7 Feb 2015 09:58:06 +0000 (10:58 +0100)]
drm: Sync drm_context.c with Linux 3.10
François Tigeot [Sat, 7 Feb 2015 09:37:27 +0000 (10:37 +0100)]
idr: Implement idr_alloc()
Sepherosa Ziehau [Sat, 7 Feb 2015 14:38:52 +0000 (22:38 +0800)]
iwn: Release WLAN serializer before ieee80211_ifattach()
Mainly to avoid dead-locking domsg to netisrs in ieee80211_ifattach()
Reported-by: Rolinh on IRC
Sepherosa Ziehau [Fri, 6 Feb 2015 13:12:05 +0000 (21:12 +0800)]
ecc/e5: Strip comma from device description
Sepherosa Ziehau [Fri, 6 Feb 2015 13:06:13 +0000 (21:06 +0800)]
ecc/e5: Declare dimmmtr near its usage
Sepherosa Ziehau [Fri, 6 Feb 2015 13:00:34 +0000 (21:00 +0800)]
ecc/e5: No need to check rank disabled bits
Sepherosa Ziehau [Fri, 6 Feb 2015 12:49:06 +0000 (20:49 +0800)]
memtemp: Add Intel E5 v2 memory thermal sensor support
François Tigeot [Fri, 6 Feb 2015 18:05:49 +0000 (19:05 +0100)]
Sync intel_sdvo.c with Linux 3.10
Tomohiro Kusumi [Mon, 26 Jan 2015 18:20:38 +0000 (03:20 +0900)]
lib/libhammer: don't hide sysctlbyname(3) results
This patch makes libhammer_btree_stats() and libhammer_io_stats() check
return value of each sysctlbyname(3) instead of only returning the last
result.
Closes: #2779
Tomohiro Kusumi [Sat, 31 Jan 2015 21:35:10 +0000 (06:35 +0900)]
hammer: fix terminology of "large block"
This cleanup patch changes terminology "large block" to "big block".
- Both "large block" and "big block" are widely used in hammer source from
kernel to userspace, however these two refer to the same data structure which
is a 8MB sized chunk within low level blockmapped storage layer.
- The original design document https://www.dragonflybsd.org/hammer/hammer.pdf
uses big block for this data structure. Having two expressions in its
implementation is confusing and makes grep difficult.
Closes: #2782
Antonio Huete Jimenez [Sat, 7 Feb 2015 11:04:04 +0000 (12:04 +0100)]
libhammer - Fix a fd leak on the previous commit
- Also check fd just in case it's invalid
Submitted-by: tkusumi (with modification)
Closes: #2784
Matthew Dillon [Sat, 7 Feb 2015 06:16:40 +0000 (22:16 -0800)]
libc - Disallow an excessively large pattern space
* Disallow pattern spaces which would cause intermediate calculations
to overflow size_t.
* Since the pattern length must be in the hundreds of megabytes to trigger
the issue on a 32-bit machine (and something insane on a 64-bit machine),
it is unlikely that any software is at risk. But still a good idea to
deal with the case.
CERT: VU#695940
Sascha Wildner [Sat, 7 Feb 2015 04:23:09 +0000 (05:23 +0100)]
Remove no longer used posix4(9) manual page.
François Tigeot [Fri, 6 Feb 2015 16:48:41 +0000 (17:48 +0100)]
drm/i915: Remove intel_encoder_noop and intel_crtc_noop()
Reducing differences with Linux 3.10
François Tigeot [Fri, 6 Feb 2015 16:23:36 +0000 (17:23 +0100)]
drm: Sync drm_buffer.c with Linux 3.10
François Tigeot [Fri, 6 Feb 2015 16:14:47 +0000 (17:14 +0100)]
drm: Sync drm_crtc.c and drm_crtc_helper.c with Linux 3.10
François Tigeot [Fri, 6 Feb 2015 15:50:03 +0000 (16:50 +0100)]
drm/i915: Use kzalloc()
Fix little differences with Linux 3.10 when appropriate.
François Tigeot [Fri, 6 Feb 2015 15:44:33 +0000 (16:44 +0100)]
drm: Implement kzalloc()
Sepherosa Ziehau [Fri, 6 Feb 2015 06:27:28 +0000 (14:27 +0800)]
acpica: Fix
3abee989bb021ca7a77508792e45828ce9d53b28
STDERR_FILENO instead of magic number 1 should be used.
Sepherosa Ziehau [Fri, 6 Feb 2015 05:17:06 +0000 (13:17 +0800)]
acpica: Unbreak iasl compile
Sepherosa Ziehau [Thu, 5 Feb 2015 13:17:27 +0000 (21:17 +0800)]
wlan: Destroy main ifp until all vaps are destroyed
The main ifp is referenced by ieee80211_vap_destroy(), so it must
be destroyed _after_ all vaps are destroyed.
Tested-by: dillon@
Sepherosa Ziehau [Thu, 5 Feb 2015 13:16:52 +0000 (21:16 +0800)]
em/emx: Apply the same TSO workaround as 82571/82572 for i217/i218
Tested-by: dillon@ on i217
Sepherosa Ziehau [Thu, 5 Feb 2015 14:38:12 +0000 (22:38 +0800)]
ecc.4: Adjust according to the recent changes
Antonio Huete Jimenez [Thu, 5 Feb 2015 23:29:55 +0000 (00:29 +0100)]
procfs - Properly return error codes for ioctl calls
Sascha Wildner [Thu, 5 Feb 2015 22:49:01 +0000 (23:49 +0100)]
kern_clock.c: Clarify that my last commit msg meant to say 'time_uptime'.
Also fix some typos in the comments.
Sascha Wildner [Thu, 5 Feb 2015 18:59:20 +0000 (19:59 +0100)]
Sync ACPICA with Intel's version
20150204.
* A raw gpe handling mechanism was created to allow better handling of
GPE storms that aren't easily managed by the normal handler. The raw
handler allows disabling/renabling of the the GPE so that interrupt
storms can be avoided in cases where events cannot be timely serviced.
* Always modify GPE registers under the GPE lock.
* Various small bug fixes.
* Update copyrights.
Antonio Huete Jimenez [Thu, 5 Feb 2015 15:28:14 +0000 (16:28 +0100)]
libhammer - Change the way PFS mounts are detected
- libhammer_find_pfs_mount() now takes a single argument which
is the unique_uuid used for the search.
- This solves a problem when passing the path of a PFS to commands
like 'hammer info' which mistakenly took the argument as the PFS#0.
- Adjust manpage to reflect current status.
Sascha Wildner [Thu, 5 Feb 2015 21:27:24 +0000 (22:27 +0100)]
kernel/clock: Also document that time_second is since booting.
Matthew Dillon [Thu, 5 Feb 2015 20:27:55 +0000 (12:27 -0800)]
kernel - Fix documentation for time_second
* time_second was improperly documented as an uptime. It is not.
It is an approximate real time and might not be monotonic.
time_uptime should be used if an approximate monotonic uptime is
desired.
Sascha Wildner [Thu, 5 Feb 2015 08:43:28 +0000 (09:43 +0100)]
kernel/wlan: Remove an unused malloc type.
Sepherosa Ziehau [Wed, 4 Feb 2015 14:01:19 +0000 (22:01 +0800)]
wlan: Make sure that WLAN serializer is not held for if_{detach,attach}()
Tested-by: dillon@
Sepherosa Ziehau [Wed, 4 Feb 2015 12:40:26 +0000 (20:40 +0800)]
ecc: Support Intel E5 v2 memory controller
Sepherosa Ziehau [Wed, 4 Feb 2015 02:44:59 +0000 (10:44 +0800)]
build: Unbreak GENERIC building; LINT seems fine.
Sepherosa Ziehau [Tue, 3 Feb 2015 12:42:39 +0000 (20:42 +0800)]
x86_64: Disable mwait extention if there is no sub-states for C1
DragonFly-bug: http://bugs.dragonflybsd.org/issues/2781
Sascha Wildner [Tue, 3 Feb 2015 22:14:01 +0000 (23:14 +0100)]
kernel/iwn: Comment out IWN_DEBUG in the Makefile.
Sascha Wildner [Tue, 3 Feb 2015 22:11:25 +0000 (23:11 +0100)]
kernel/wlan: Re-fix printf issues in debug messages.
Sascha Wildner [Tue, 3 Feb 2015 22:20:55 +0000 (23:20 +0100)]
kernel/ath: Disallow overriding ATH_RXBUF and ATH_TXBUF.
They are now forced to 512.
Sascha Wildner [Tue, 3 Feb 2015 21:56:02 +0000 (22:56 +0100)]
kernel/ath: Fix building with AH_DEBUG.
Sascha Wildner [Tue, 3 Feb 2015 21:29:04 +0000 (22:29 +0100)]
iwn/wpi: Fix building firmwares into the kernel.
Johannes Hofmann [Fri, 30 Jan 2015 19:30:11 +0000 (20:30 +0100)]
kernel/iwn - Fix MAC address printing in verbose mode
* Fix MAC address printing in verbose mode after report
from FreeBSD
Johannes Hofmann [Thu, 29 Jan 2015 22:27:18 +0000 (23:27 +0100)]
kernel/wpi - Add back ASSERT_ALTQ_SQ_DEFAULT
* Add back ASSERT_ALTQ_SQ_DEFAULT to wpi_start()
Johannes Hofmann [Thu, 29 Jan 2015 22:19:25 +0000 (23:19 +0100)]
kernel/wpi - Use IFNET_STAT_INC macros
* Switch back to use IFNET_STAT_INC macros
Johannes Hofmann [Thu, 29 Jan 2015 22:00:32 +0000 (23:00 +0100)]
kernel/wpi - Update wpi firmware to iwlwifi-3945-15.32.2.9
* Update wpi firmware to iwlwifi-3945-15.32.2.9
Taken-from: FreeBSD
Johannes Hofmann [Thu, 29 Jan 2015 21:55:08 +0000 (22:55 +0100)]
kernel/wpi - Report wpi from FreeBSD
* Report wpi from FreeBSD as of FreeBSD git revision
d692929f2
Matthew Dillon [Tue, 20 Jan 2015 23:51:44 +0000 (15:51 -0800)]
wlan - update README.DRAGONFLY for driver porting
* Update README.DRAGONFLY to describe in detail how to port
low level wifi drivers.
Matthew Dillon [Tue, 20 Jan 2015 22:22:30 +0000 (14:22 -0800)]
kernel/ath - export athdev_printf()
* Fix the export of athdev_printf() so it is unconditional.
Matthew Dillon [Tue, 20 Jan 2015 22:21:04 +0000 (14:21 -0800)]
wlan - export ieee80211_add_ssid()
* Export ieee80211_add_ssid() so iwn can use it.
Matthew Dillon [Tue, 20 Jan 2015 22:20:09 +0000 (14:20 -0800)]
kernel/iwn - Report iwn from FreeBSD
* Report iwn from FreeBSD
Matthew Dillon [Tue, 20 Jan 2015 16:51:00 +0000 (08:51 -0800)]
kernel/ath - Synchronize with FreeBSD
* Synchronize ath with FreeBSD.
Matthew Dillon [Sun, 11 Jan 2015 20:12:07 +0000 (12:12 -0800)]
wlan - Finish ieee80211_realign()
* Finish implementing ieee80211_realign() which is now being called
in more paths, particularly IWN.
* Fixes IWN.
Matthew Dillon [Sun, 11 Jan 2015 17:50:12 +0000 (09:50 -0800)]
wlan - Update wlan from Adrian / FreeBSD
* Update the wlan infrastructure, initially working with ath.
Markus Pfeiffer [Mon, 2 Feb 2015 06:57:55 +0000 (06:57 +0000)]
usb4bsd: Re-port the if_rum driver
Matthew Dillon [Tue, 3 Feb 2015 17:14:59 +0000 (09:14 -0800)]
if_bridge - Fix man page
* Man page incorrectly said MAC cache was 1024 entries. It is 4096.
Reported-by: marino
Matthew Dillon [Tue, 3 Feb 2015 03:33:02 +0000 (19:33 -0800)]
if_bridge - Increase MAC cache maximum to 4096
* Increase the default maximum number of records the bridge can keep track
of from 100 to 4096, which is more contemporary with what hardware
bridges deal with.
Matthew Dillon [Tue, 3 Feb 2015 03:28:32 +0000 (19:28 -0800)]
if_bridge - Fix issue where source MAC is not recorded.
* The bridge has to record which interface the source MAC came from
so transmissions back to that MAC can be unicast. Otherwise transmissions
will be broadcast to all bridge elements.
The receive packet path case where a packet is sent to the bridge's own
MAC address (terminating on the bridge host) from one of the member
interfaces was not recording the MAC, resulting in duplicate packets
(most obvious when testing with 'ping').
* Fix the code path in question so the MAC is properly recorded.
Imre Vadasz [Mon, 2 Feb 2015 00:39:00 +0000 (01:39 +0100)]
libc -- dmalloc: Fix uninitialized variable in memrealloc().
Sascha Wildner [Sun, 1 Feb 2015 14:26:50 +0000 (15:26 +0100)]
libc/db: Fix build with HASH_STATISTICS defined.
Sascha Wildner [Sun, 1 Feb 2015 13:47:37 +0000 (14:47 +0100)]
libc/getifaddrs: Remove dead code.
We have NET_RT_IFLIST and ifm_data, but we don't have the BSDi specific
ifam_data.
François Tigeot [Sun, 25 Jan 2015 18:33:18 +0000 (19:33 +0100)]
drm/i915: Update to Linux 3.10
* Page flip improvements, increased robustness after GPU hangs.
Races have been fixed in the vblank and pageflip completion
interrupt handlers of GEN2-GEN4 GPUs.
* Hotplug IRQ storm detection and mitigation.
* Various bug fixes and performance improvements from GEN4 to
Haswell GPUs.
Less "Unclaimed register" messages should be visible on Haswell.
* Improved support for Valleyview / Baytrail GPUs.
* Better overclocking support from Sandy Bridge to Haswell.
Improved frequency scaling (turbo) on Haswell.
* Improved display detection and modesetting. It is now possible to
precompute possible desired display pipe configurations without
changing the hardware state.
* Daniel Vetter's blog contains more detailed information:
http://blog.ffwll.ch/2013/04/neat-drmi915-stuff-for-310.html
Sascha Wildner [Sat, 31 Jan 2015 23:38:08 +0000 (00:38 +0100)]
Sync zoneinfo database with tzdata2015a from ftp://ftp.iana.org/tz/releases
Changes affecting future time stamps
* The Mexican state of Quintana Roo, represented by America/Cancun,
will shift from Central Time with DST to Eastern Time without DST
on 2015-02-01 at 02:00. (Thanks to Steffen Thorsen and Gwillim Law.)
* Chile will not change clocks in April or thereafter; its new standard time
will be its old daylight saving time. This affects America/Santiago,
Pacific/Easter, and Antarctica/Palmer. (Thanks to Juan Correa.)
* New leap second 2015-06-30 23:59:60 UTC as per IERS Bulletin C 49.
(Thanks to Tim Parenti.)
Changes affecting past time stamps
* Iceland observed DST in 1919 and 1921, and its 1939 fallback
transition was Oct. 29, not Nov. 29. Remove incorrect data from
Shanks about time in Iceland between 1837 and 1908.
* Some more zones have been turned into links, when they differed
from existing zones only for older time stamps. As usual,
these changes affect UTC offsets in pre-1970 time stamps only.
The affected zones are: Asia/Aden, Asia/Bahrain, Asia/Kuwait,
and Asia/Muscat.
Changes affecting commentary
* The leapseconds file commentary now mentions the expiration date.
(Problem reported by Martin Burnicki.)
* Update Mexican Library of Congress URL.
Robin Hahling [Sat, 31 Jan 2015 10:06:38 +0000 (11:06 +0100)]
libcrypto: Raise WARNS to 2 and remove HAVE_CRYPTODEV definition.
Add DragonFly as supporting cryptodev to vendor sources instead of
using HAVE_CRYPTODEV for it (which led to a redefinition warning).
This patch will be submitted upstream.
Matthew Dillon [Sat, 31 Jan 2015 19:16:37 +0000 (11:16 -0800)]
kernel - Remove if_re debugging message
* Remove the 'Debug: OACTIVE was not set when re_tx_free was below minimum!'
kprintf. The case is known to happen every once in a while, we no longer
need to validate that the path is being hit.
Reported-by: Tim Darby
Matthew Dillon [Sat, 31 Jan 2015 19:03:01 +0000 (11:03 -0800)]
kernel - Add sysctl to loosen ND packet restrictions
* Add a sysctl to allow ND packet restrictions to be loosened,
default is off. A (hopefully very) few IPV6 configurations might need
it.
* Use of this sysctl is NOT recommended as it opens up multiple avenues
for messing up IPV6 routing.
Submitted-by: gpr PR/2780
Taken-From: FreeBSD
Sascha Wildner [Sat, 31 Jan 2015 10:59:20 +0000 (11:59 +0100)]
<sys/resource.h>: Add RLIM_SAVED_{CUR,MAX} and separate BSD specific stuff.
Sascha Wildner [Sat, 31 Jan 2015 10:58:41 +0000 (11:58 +0100)]
<sys/resource.h>: Adjust whitespace (no functional change).
Matthew Dillon [Fri, 30 Jan 2015 02:26:34 +0000 (18:26 -0800)]
hammer2 - fix I/O race, flesh out bulkfree
* Fix an I/O race which could cause hammer2 to block on an iocb forever.
A TAILQ was being tested outside of a spin-lock and could race the
related atomic op on the flag.
* Add signal checking for the bulkfree scan, allowing it to be interrupted.
* Loop sbase/sstop to cover the whole volume. The in-memory bitmap must
be bzero()d on each loop.
Matthew Dillon [Thu, 29 Jan 2015 18:21:46 +0000 (10:21 -0800)]
hammer2 - work on bmap->avail tracking, fix DIO issue w/block reuse
* Start working on tracking bmap->avail in a way that the bulkfree
code can undo.
* DIO was allowing mixed 16/64K buffers, but bulkfree doesn't invalidate
cached DIOs and the re-use case might want a larger buffer. For now
deal with it by just having DIO use 64K buffers.
Sascha Wildner [Wed, 28 Jan 2015 21:02:46 +0000 (22:02 +0100)]
UPDATING: Mention recently removed kernel options.
Sascha Wildner [Wed, 28 Jan 2015 20:58:53 +0000 (21:58 +0100)]
Remove <sys/_posix.h>. It serves no purpose anymore.
Sascha Wildner [Wed, 28 Jan 2015 20:56:43 +0000 (21:56 +0100)]
kernel: Retire the P1003_1B and _KPOSIX_VERSION kernel options.
Sascha Wildner [Wed, 28 Jan 2015 19:25:08 +0000 (20:25 +0100)]
<sys/semaphore.h>: Remove dead p1003_1b code.
Sascha Wildner [Wed, 28 Jan 2015 19:27:02 +0000 (20:27 +0100)]
<sys/sysctl.h>: Remove <sys/_posix.h> include. It is not needed here.
Sascha Wildner [Wed, 28 Jan 2015 20:20:34 +0000 (21:20 +0100)]
kernel/mib: Remove <sys/_posix.h> dependency.
This commit bumps the value of the kern.posix1version sysctl from 199309 to
200112. I know no port which checks this sysctl.
Imre Vadasz [Wed, 28 Jan 2015 19:32:18 +0000 (20:32 +0100)]
kernel/snd_hda: Enable uncacheable DMA attribute as default again.
* Set the HDAC_F_DMA_NOCACHE flag as default (Reinserts a line that was
removed in commit
403fad76b1ecab3ce8ef68b48445be3a9b0a4e74).
* Allocate at least PAGE_SIZE aligned memory in hdac_dma_alloc() when
HDAC_F_DMA_NOCACHE is set.
Sascha Wildner [Wed, 28 Jan 2015 19:11:20 +0000 (20:11 +0100)]
<sys/mman.h>: Some cleanup.
* Handle _P1003_1B_VISIBLE and _POSIX_SOURCE checks with __POSIX_VISIBLE.
* Separate BSD specific defines/prototypes better.
Sascha Wildner [Mon, 26 Jan 2015 19:19:57 +0000 (20:19 +0100)]
<sys/mman.h>: Whitespace cleanup (no functional change).
John Marino [Wed, 28 Jan 2015 11:47:55 +0000 (12:47 +0100)]
bsd.hostlib.mk: Fix case where SRCS is defined but empty
This is similar to the recent changes to bsd.lib.mk and bsd.prog.mk.
The check covers two augmentations to OBJS, plus let's add some indents
for readability. Note the second augmentation didn't have a bug in it,
so including it inside the wrap only saves having to evaluate it.
Sepherosa Ziehau [Sat, 24 Jan 2015 05:42:14 +0000 (13:42 +0800)]
inet6: Dispatch nd6_timer to netisr0 to run
This function now accesses ifindex2ifnet global variable in netisr0.
Tomohiro Kusumi [Sun, 25 Jan 2015 14:13:04 +0000 (23:13 +0900)]
lib/libhammer: cleanups
Using pointer dereference instead of several mntbuf[curmount].f_xxx
makes code clear and also makes lines fit within 80 characters.
Tomohiro Kusumi [Sun, 25 Jan 2015 13:41:59 +0000 (22:41 +0900)]
lib/libhammer: cleanups
This patch has some minor cleanups of libhammer.
hunk1: Remove parentheses
hunk2: Remove parentheses
hunk3: Remove forward declaration of obsolete function.
find_pfs_mount() was re-implemented as a part of libhammer
and commit
4f09feab replaces it with libhammer version.
Tomohiro Kusumi [Sun, 25 Jan 2015 13:38:50 +0000 (22:38 +0900)]
lib/libhammer: fix minor memory leaks
This patch fixes minor memory leaks of libhammer.
hunk1: This isn't necessary since it never gets here.
_libhammer_malloc() exit(1) if it failed to calloc().
hunk2: Add missing close(fd).
hunk3: Add missing free(mntbuf).
Tomohiro Kusumi [Sat, 24 Jan 2015 15:55:21 +0000 (00:55 +0900)]
hammer: some trivial fixes
This patch fixes some trivial things.
hunk1: Cleanup incomplete sentence
hunk2: Sync comment with changes made in
4a2796f3
hunk3: Use right function name
hunk4: Add missing comment
hunk5: Add missing \n
Closes: #2777
John Marino [Tue, 27 Jan 2015 22:20:44 +0000 (23:20 +0100)]
bsd.*.mk: Fix bug when SRCS is explicitly defined empty
There exists a valid possibility of setting a makefile to build a program
or library exclusively with OBJS defined. In the case of bsd.prog.mk,
not defining SRCS will cause the definition of OBJS to be overwritten
with ${PROG}.c or ${PROG}.cc. Defining SRCS as an empty list will stop
OBJS from being overwritten, but there's a bug. It adds ".o" to the end
of the list.
For example:
======================
OBJS= first.o second.o
SRCS=
.include <bsd.prog.mk>
======================
make -V OBJS
first.o second.o .o
This is caused by a transformation assuming SRCS has at least one item
in the list. It's fixed by only performing the transformation if SRCS
is not empty. I'm adding this check to both bsd.prog.mk and bsd.lib.mk
Sascha Wildner [Tue, 27 Jan 2015 21:54:57 +0000 (22:54 +0100)]
libc/sigaltstack: Return EPERM upon disabling the active stack, not EINVAL.
Per POSIX, FreeBSD and OpenBSD agree too. NetBSD doesn't, hmmm.
Imre Vadasz [Tue, 27 Jan 2015 18:36:16 +0000 (19:36 +0100)]
kernel/sound: Always create the devfs aliases when loading a sound driver.
Sascha Wildner [Tue, 27 Jan 2015 09:28:49 +0000 (10:28 +0100)]
kernel/snd_hda: Fix module unloading.
Submitted-by: ivadasz
Sascha Wildner [Tue, 27 Jan 2015 09:27:59 +0000 (10:27 +0100)]
kernel/snd_hda: Re-apply
6d233c5d040778c2655923f2a20676e7e519ceca.
It still occurs for us.
Pointed-out-by: ivadasz
Sascha Wildner [Tue, 27 Jan 2015 08:12:20 +0000 (09:12 +0100)]
Clarify message about 'make rescue' after installworld.
Robin Hahling [Tue, 27 Jan 2015 05:56:42 +0000 (06:56 +0100)]
openssh: Adjust DragonFly version to actual openssh p1 release date.
Sascha Wildner [Tue, 27 Jan 2015 05:31:19 +0000 (06:31 +0100)]
libusb: Sync with FreeBSD.
Matthew Dillon [Tue, 27 Jan 2015 02:07:05 +0000 (18:07 -0800)]
hammer2 - Adjust TODO
* Add bulkfree TODO items.
Matthew Dillon [Tue, 27 Jan 2015 02:04:30 +0000 (18:04 -0800)]
hammer2 - Add bulkfree directive.
* Add a bulkfree directive for testing the bulkfree code.
Matthew Dillon [Tue, 27 Jan 2015 02:00:53 +0000 (18:00 -0800)]
hammer2 - bulkfree work 2/many
* Compare the in-memory bitmap against the on-media bitmap and issue
transitions.
in-memory 00 - 11 (allocated) -> 10 (staged)
in-memory 00 - 10 (staged) -> 00 (free)
in-memory 11 - 10 (staged) -> 11 (allocated) (handles live races)
Matthew Dillon [Mon, 26 Jan 2015 23:48:03 +0000 (15:48 -0800)]
hammer2 - Stabilize new I/O infrastructure and work on bulkfree
* Stabilize the new asynchronous hammer_io infrastructure.
* Document the new simpler bulkfree scan algorithm. The new algorithm
scans meta-data in two passes, rebuilding the freemap in kernel memory.
It then compares the map against the live media.
The first pass transitions the freemap from (11 = allocated) to
(10 = staged for deletion), the second pass transitions (10) either back
to (11) if the block is found to not actually be freeable or to (00) if
the block winds up being actually freeable.
The two pass mechanism deals with any races and allows the scan to occur
essentially unlocked,
* Start work on the bulkfree scan algorithm.
Matthew Dillon [Mon, 26 Jan 2015 23:46:25 +0000 (15:46 -0800)]
kernel - Add swap-backed kernel memory allocator API
* Add kmem_alloc_swapbacked() and kmem_free_swapbacked(). Allocates kernel
memory which is pageable and backed by swap.
Matthew Dillon [Mon, 26 Jan 2015 23:45:11 +0000 (15:45 -0800)]
kernel - limit backtrace-on-panic to 6 call levels
* Limit the backtrace-on-panic to 6 call levels to try to keep more
information leading up to the crash (such as the panic message) on
row-limited consoles.