Rui Paulo [Sat, 8 Jun 2013 16:02:31 +0000 (16:02 +0000)]
Import Kevin Lo's port of urtwn(4) from OpenBSD. urtwn(4) is a driver for the
Realtek RTL8188CU/RTL8192CU USB IEEE 802.11b/g/n wireless cards.
This driver requires microcode which is available in FreeBSD ports:
net/urtwn-firmware-kmod.
Hiren ported the urtwn(4) man page from OpenBSD and Glen just commited a port
for the firmware.
TODO:
- 802.11n support
- Stability fixes - the driver can sustain lots of traffic but has trouble
coping with simultaneous iperf sessions.
- fix debugging
Gleb Smirnoff [Sat, 8 Jun 2013 13:27:57 +0000 (13:27 +0000)]
Add new system call - aio_mlock(). The name speaks for itself. It allows
to perform the mlock(2) operation, which can consume a lot of time, under
control of aio(4).
Reviewed by: kib, jilles
Sponsored by: Nginx, Inc.
Ed Schouten [Sat, 8 Jun 2013 13:19:11 +0000 (13:19 +0000)]
Add proper __sync_*() intrinsics for MIPS.
To make <stdatomic.h> work on MIPS (and ARM) using GCC, we need to
provide implementations of the __sync_*() functions. I already added
these functions for 4 and 8 byte types to libcompiler-rt some time ago,
based on top of <machine/atomic.h>.
Unfortunately, <machine/atomic.h> only provides a subset of the features
needed to implement <stdatomic.h>. This means that in some cases we had
to do compare-and-exchange calls in loops, where a simple ll/sc would
suffice.
Also implement these functions for 1 and 2 byte types. MIPS only
provides ll/sc instructions for 4 and 8 byte types, but this is of
course no limitation. We can simply load 4 bytes and use some bitmask
tricks to modify only the bytes affected.
ZFS should proactively evict freed blocks from the cache.
Even though these freed blocks will never be used again, and thus
will eventually be evicted, this causes us to use memory
inefficiently for 2 reasons:
1. A block that is freed has no chance of being accessed again, but
will be kept in memory preferentially to a block that was accessed
before it (and is thus older) but has not been freed and thus has
at least some chance of being accessed again.
2. We partition the ARC into several buckets:
user data that has been accessed only once (MRU)
metadata that has been accessed only once (MRU)
user data that has been accessed more than once (MFU)
metadata that has been accessed more than once (MFU)
The user data vs metadata split is somewhat arbitrary, and the
primary control on how much memory is used to cache data vs metadata
is to simply try to keep the proportion the same as it has been in the
past (each bucket "evicts against" itself). The secondary control is
to evict data before evicting metadata.
Because of this bucketing, we may end up with one bucket mostly
containing freed blocks that are very old, while another bucket has
more recently accessed, still-allocated blocks. Data in the useful
bucket (with still-allocated blocks) may be evicted in preference to
data in the useless bucket (with old, freed blocks).
On dcenter, we saw that the MFU metadata bucket was 230MB, while the
MFU data bucket was 27GB and the MRU metadata bucket was 256GB.
However, the vast majority of data in the MRU metadata bucket (256GB)
was freed blocks, and thus useless. Meanwhile, the MFU metadata bucket
(230MB) was constantly evicting useful blocks that will be soon needed.
The problem of cache segmentation is a larger problem that needs more
investigation. However, if we stop caching freed blocks, it should
reduce the impact of this more fundamental issue.
Andrew Turner [Sat, 8 Jun 2013 07:16:22 +0000 (07:16 +0000)]
Merge in changes from NetBSD:
* Remove support for non-elf files.
* Add the VFP setjmp magic numbers.
* Add the offsets for the VFP registers within the buffer.
1) Only multi-TD isochronous transfers should use NORMAL
type after specific type as per XHCI specification.
2) BEI bit is only available in NORMAL and ISOCHRONOUS
TRB types. Don't use this bit for other types to avoid
hardware asserts. Reserved bits should be don't care
though ...
Andrew Turner [Fri, 7 Jun 2013 22:01:06 +0000 (22:01 +0000)]
Finish pulling in the NetBSD setjmp/longjmp updates on ARM.
Store/restore the VFP registers in setjmp/longjmp on ARM EABI if VFP is
enabled in the kernel. It checks the hw.floatingpoint sysctl to see if
floating-point is available and uses this to determine if it should store
them. If it does it uses a different magic value so longjmp is able to know
if it should load them.
Ed Maste [Fri, 7 Jun 2013 21:40:02 +0000 (21:40 +0000)]
Add a new knob WITH_DEBUG_FILES to control the building of standalone
debug files for userland programs and libraries. The "-g" debug flag
is automatically applied when WITH_DEBUG_FILES is set.
The debug files are now named ${prog}.debug and ${shlib}.debug for
consistency with other systems and documentation. In addition they are
installed under /usr/lib/debug, to simplify the process of installing
them if needed after a crash. Users of bsd.{prog,lib}.mk outside of the
base system place the standalone debug files in a .debug subdirectory.
GDB automatically searches both of these directories for standalone
debug files.
Thanks to everyone who contributed changes, review, and testing during
development.
Andrew Turner [Fri, 7 Jun 2013 21:23:11 +0000 (21:23 +0000)]
Reduce the difference to NetBSD.
* Stop pretending we support anything other than ELF by removing code
surrounded by #ifdef __ELF__ ... #endif.
* Remove _JB_MAGIC_SETJMP and _JB_MAGIC__SETJMP, they are defined in
setjmp.h, which is able to be included from asm.
* Fix the spelling of dependent.
* Rename END _END and add END and ASEND to complement ENTRY and ASENTRY
respectively
* Add macros to simplify accessing the Global Offset Table, some of these
will be used in the upcoming update to the setjmp functions.
Add support for polling the XHCI interrupt handler when
the regular interrupt handler is not working properly or
in case of MSI interrupts which are not yet supported.
Remove interrupt setup code for FreeBSD versions older
than 700031.
Ed Maste [Fri, 7 Jun 2013 13:45:58 +0000 (13:45 +0000)]
Add libusb_get_port_numbers
libusbx deprecated libusb_get_port_path and replaced it with
libusb_get_port_numbers. The latter omits an extra parameter which was
unused in the FreeBSD implementation anyway.
Adrian Chadd [Fri, 7 Jun 2013 09:02:02 +0000 (09:02 +0000)]
Bring over the initial static bluetooth coexistence configuration
for the WB195 combo NIC - an AR9285 w/ an AR3011 USB bluetooth NIC.
The AR3011 is wired up using a 3-wire coexistence scheme to the AR9285.
The code in if_ath_btcoex.c sets up the initial hardware mapping
and coexistence configuration. There's nothing special about it -
it's static; it doesn't try to configure bluetooth / MAC traffic priorities
or try to figure out what's actually going on. It's enough to stop basic
bluetooth traffic from causing traffic stalls and diassociation from
the wireless network.
To use this code, you must have the above NIC. No, it won't work
for the AR9287+AR3012, nor the AR9485, AR9462 or AR955x combo cards.
Then you set a kernel hint before boot or before kldload, where 'X'
is the unit number of your AR9285 NIC:
# kenv hint.ath.X.btcoex_profile=wb195
This will then appear in your boot messages:
[100482] athX: Enabling WB195 BTCOEX
This code is going to evolve pretty quickly (well, depending upon my
spare time) so don't assume the btcoex API is going to stay stable.
In order to use the bluetooth side, you must also load in firmware using
ath3kfw and the binary firmware file (ath3k-1.fw in my case.)
Tested:
* AR9280, no interference
* WB195 - AR9285 + AR3011 combo; STA mode; basic bluetooth inquiries
were enough to cause traffic stalls and disassociations. This has
stopped with the btcoex profile code.
TODO:
* Importantly - the AR9285 needs ASPM disabled if bluetooth coexistence
is enabled. No, I don't know why. It's likely some kind of bug to do
with the AR3011 sending bluetooth coexistence signals whilst the device
is asleep. Since we don't actually sleep the MAC just yet, it shouldn't
be a problem. That said, to be totally correct:
+ ASPM should be disabled - upon attach and wakeup
+ The PCIe powersave HAL code should never be called
Look at what the ath9k driver does for inspiration.
* Add WB197 (AR9287+AR3012) support
* Add support for the AR9485, which is another combo like the AR9285
* The later NICs have a different signaling mechanism between the MAC
and the bluetooth device; I haven't even begun to experiment with
making that HAL code work. But it should be a lot more automatic.
* The hardware can do much more interesting traffic weighting with
bluetooth and wifi traffic. None of this is currently used.
Ideally someone would code up something to watch the bluetooth traffic
GPIO (via an interrupt) and then watch it go high/low; then figure out
what the bluetooth traffic is and adjust things appropriately.
* If I get the time I may add in some code to at least track this stuff
and expose statistics. But it's up to someone else to experiment with
the bluetooth coexistence support and add the interesting stuff (like
"real" detection of bulk, audio, etc bluetooth traffic patterns and
change wifi parameters appropriately - eg, maximum aggregate length,
transmit power, using quiet time to control TX duty cycle, etc.)
Use getnameinfo(3) instead of inet_ntop(3) to make printable versions of
sockaddr_in6 structures. getnameinfo(3) does the same thing, but it is
also able to represent a scope zone id as described in the RFC 4007.
Jaakko Heinonen [Fri, 7 Jun 2013 07:17:46 +0000 (07:17 +0000)]
Revert r238399.
The "failok" option doesn't have any effect at all unless specified in
fstab(5) and combined with the -a flag. The "failok" option is already
documented in fstab(5).
Adrian Chadd [Fri, 7 Jun 2013 05:17:58 +0000 (05:17 +0000)]
Add bluetooth fixes to the AR5416/AR92xx HAL:
* Call the bluetooth setup function during the reset path, so the bluetooth
settings are actually initialised.
* Call the AR9285 diversity functions during bluetooth setup; so the AR9285
diversity and antenna configuration registers are correctly programmed
* Misc debugging info.
Tested:
* AR9285+AR3011 bluetooth combo; this code itself doesn't enable bluetooth
coexistence but it's part of what I'm currently using.
Pyun YongHyeon [Fri, 7 Jun 2013 01:21:09 +0000 (01:21 +0000)]
Correct setting TX random backoff register. This register is
implemented as a 10 bits linear feedback shift register so only
lower 10 bits are valid.
Because this register is used to initialize random backoff interval
register only when resolved duplex is half-duplex, it wouldn't have
caused issues in these days.
Pyun YongHyeon [Fri, 7 Jun 2013 01:01:39 +0000 (01:01 +0000)]
Do not report current link status if driver is not running.
Reporting link status in driver has a side-effect that makes mii(4)
check current link status. mii(4) will call link status change
callback when it sees link state change. Normally this wouldn't
have problems. However, ASF/IPMI firmware can actively access PHY
regardless of driver's running state such that reporting link
status for not-running interface can generate meaningless link
UP/DOWN messages.
This change also makes dhclient think driver got a valid link
regardless of link establishment so it will bypass dhclient's
initial link status check. I think that wouldn't be issue
though.
Xin LI [Thu, 6 Jun 2013 23:21:41 +0000 (23:21 +0000)]
MFV r251474:
* Illumos zfs issue #3137 L2ARC compression
Whether or not to compress buffers entering the L2ARC is
controlled by "compression" setting on the dataset, when
compression is not "off", L2ARC compression is enabled.
The compress method is always LZ4 for L2ARC when enabled
because it works best for the scenario.
Attilio Rao [Thu, 6 Jun 2013 18:19:26 +0000 (18:19 +0000)]
Complete r251452:
Avoid to busy/unbusy a page in cases where there is no need to drop the
vm_obj lock, more nominally when the page is full valid after
vm_page_grab().
John Baldwin [Thu, 6 Jun 2013 14:43:19 +0000 (14:43 +0000)]
Do not compare the existing mask of a cpuset with a new mask when changing
the mask of a cpuset. Also, change the cpuset's mask before updating the
masks of all children. Previously changing a cpuset's mask first required
setting the mask to a super-set of both the old and new masks and then
changing it a second time to the new mask.
Jeff Roberson [Wed, 5 Jun 2013 23:53:00 +0000 (23:53 +0000)]
- Consolidate duplicate code into support functions.
- Split the bqlock into bqclean and bqdirty locks.
- Only acquire the wakeup synchronization locks when we cross a
threshold requiring them.
- Restructure the way flushbufqueues() targets work so they are more
smp friendly and sane.
Rick Macklem [Wed, 5 Jun 2013 23:17:39 +0000 (23:17 +0000)]
Add a "-v" (verbose) option to the gssd daemon, to help
with diagnosis of kerberized NFS mount problems. When set,
messages are sent to syslog() (or fprintf(stderr,...) if
"-d" is also specified) to indicate activity/results of
kgssapi upcalls.
Adrian Chadd [Wed, 5 Jun 2013 22:23:13 +0000 (22:23 +0000)]
Enable slow diversity combining for the AR9285.
Now that I understand what's going on - and the RX antenna array maps
to what the receive LNA configuration actually is - I feel comfortable
in enabling this.
If people do have issues with this, there's enough debugging now available
that we have a chance to diagnose it without writing it up as 'weird
crap.'
Tested:
* AR9285 STA w/ diversity combining enabled in EEPROM
Adrian Chadd [Wed, 5 Jun 2013 22:21:13 +0000 (22:21 +0000)]
As a temporary work-around (read: until there's a nice API for exposing
and controlling this form of antenna diversity) - print out the AR9285
antenna diversity configuration at attach time.
This will help track down and diagose if/when people have connectivity
issues on cards (eg if they connect a single antenna to LNA1, yet the
card has RX configured to only occur on LNA2.)
Tested:
* AR9285 w/ antenna diversity enabled in EEPROM;
* AR9285 w/ antenna diversity disabled in EEPROM; mapping only to a
single antenna (LNA1.)
Dimitry Andric [Wed, 5 Jun 2013 19:46:39 +0000 (19:46 +0000)]
Pull in r183297 from upstream llvm trunk:
PR15662: Optimized debug info produces out of order function
parameters
When a function is inlined we lazily construct the variables
representing the function's parameters. After that, we add any
remaining unused parameters.
If the function doesn't use all the parameters, or uses them out of
order, then the DWARF would produce them in that order, producing a
parameter order that doesn't match the source.
This fix causes us to always keep the arg variables at the start of
the variable list & in the original order from the source.
Luigi Rizzo [Wed, 5 Jun 2013 17:27:59 +0000 (17:27 +0000)]
- fix a bug in the previous commit that was dropping the last packet
from each batch flowing on the VALE switch
- feature: add glue for 'indirect' buffers on the sender side:
if a slot has NS_INDIRECT set, the netmap buffer contains pointer(s)
to the actual userspace buffers, which are accessed with copyin().
The feature is not finalised yet, as it will likely need to deal
with some iovec variant for proper scatter/gather support.
This will save one copy for clients (e.g. qemu) that cannot
use the netmap buffer directly.
A curiosity: on amd64 copyin() appears to be 10-15% faster than pkt_copy()
or bcopy() at least for sizes of 256 and greater.
Marius Strobl [Wed, 5 Jun 2013 01:22:59 +0000 (01:22 +0000)]
- Flag sym(4) as supporting unmapped I/O; all necessary conversion actually
already has been done as part of r246713.
- Revert a part of r251402 in order to appease clang.
Adrian Chadd [Wed, 5 Jun 2013 00:45:19 +0000 (00:45 +0000)]
Implement a bit of a hack to store the AR9285/AR9485 RX LNA configuration in
the RX antenna field.
The AR9285/AR9485 use an LNA mixer to determine how to combine the signals
from the two antennas. This is encoded in the RSSI fields (ctl/ext) for
chain 2. So, let's use that here.
This maps RX antennas 0->3 to the RX mixer configuration used to
receive a frame. There's more that can be done but this is good enough
to diagnose if the hardware is doing "odd" things like trying to
receive frames on LNA2 (ie, antenna 2 or "alt" antenna) when there's
only one antenna connected.
Adrian Chadd [Wed, 5 Jun 2013 00:42:04 +0000 (00:42 +0000)]
Add a new capability flag to announce that the chip implements LNA mixing
for the RX path.
This is different to the div comb HAL flag, that says it actually
can use this for RX diversity (the "slow" diversity path implemented
but disabled in the AR9285 HAL code.)
Attilio Rao [Tue, 4 Jun 2013 22:47:01 +0000 (22:47 +0000)]
In vm_object_split(), busy and consequently unbusy the pages only when
swap_pager_copy() is invoked, otherwise there is no reason to do so.
This will eliminate the necessity to busy pages most of the times.
Alan Somers [Tue, 4 Jun 2013 22:32:33 +0000 (22:32 +0000)]
sys/dev/mps/mps.c
sys/dev/mps/mps_user.c
Fix uninitialized memory reference in mps_read_config_page. It was
referencing a field (params->hdr.Ext.ExtPageType) that would only be
set when reading an Extended config page. The symptom was that
MPSIO_READ_CFG_PAGE ioctls would randomly fail with
MPI2_IOCSTATUS_CONFIG_INVALID_PAGE errors. The solution is to
determine whether an extended or an ordinary config page is requested
by looking at the PageType field, which should be available regardless.
Similarly, mps_user_read_extcfg_header and mps_user_read_extcfg_page,
which call mps_read_config_page, had to be fixed to always set the
PageType field. They were implicitly assuming that
mps_read_config_page always operated on Extended pages.
Reviewed by: ken
Approved by: ken (mentor)
MFC after: 3 days
Marius Strobl [Tue, 4 Jun 2013 20:49:17 +0000 (20:49 +0000)]
CAM_DEV_QFREEZE handling should only be done on request submission but
neither on completion nor by SIM drivers in the first place. This issue
has been revealed by r249466.
Andrew Turner [Tue, 4 Jun 2013 19:47:26 +0000 (19:47 +0000)]
Start to merge the updated ARM NetBSD setjump/longjmp functions. To begin
with merge the functions but leave out the code to save/load the VFP
registers as that requires other changes to ensure the VFP is enabled
first.
This removes storing the old fpa registers. These were never fully
supported, and the only user of this code I can find have moved to newer
CPUs which use a VFP.
Devin Teske [Tue, 4 Jun 2013 17:08:26 +0000 (17:08 +0000)]
Shift Console->Saver menu items to place the default item at the top, making
this menu consistent with Console->Repeat, Console->Font, Console->Screenmap
and many others which place the default choice at the top.
NOTE: SVN r249751 changed things so that these menus highlight the active
selection, so the idea that the default choice (now at top) "can be selected
immediately by hitting ENTER" is not possible unless the default choice is
the currently active setting (e.g., on first visit to the menu). However, it
is still warranted to have the default choice at the top of the menu.
John Baldwin [Tue, 4 Jun 2013 15:35:37 +0000 (15:35 +0000)]
Build fix: Only <sys/cdefs.h> should be included before __FBSDID().
<sys/param.h> needs to be included after any "opt_foo.h" headers so it
sees the same set of defined macros as other headers.
Gleb Smirnoff [Tue, 4 Jun 2013 11:19:08 +0000 (11:19 +0000)]
Improve r250890, so that we stop processing of a message with zero
descriptors as early as possible, and assert that number of descriptors
is positive in unp_freerights().
Devin Teske [Tue, 4 Jun 2013 03:47:21 +0000 (03:47 +0000)]
Use f_shell_escape() instead of forking to awk. In this case, the
replacement comes with a great performance increase (as f_shell_escape()
uses the built-in based f_replaceall() which out-performs forking to
awk(1)). This should also improve readability slightly.