Commit graph

254,637 commits

Author SHA1 Message Date
Julian Anastasov
26c15cfd29 ipvs: changes related to service usecnt
Change the usage of svc usecnt during command execution:

- we check if svc is registered but we do not need to hold usecnt
reference while under __ip_vs_mutex, only the packet handling needs
it during scheduling

- change __ip_vs_service_get to __ip_vs_service_find and
__ip_vs_svc_fwm_get to __ip_vs_svc_fwm_find because now caller
will increase svc->usecnt

- put common code that calls update_service in __ip_vs_update_dest

- put common code in ip_vs_unlink_service() and use it to unregister
the service

- add comment that svc should not be accessed after ip_vs_del_service
anymore

- all IP_VS_WAIT_WHILE calls are now unified: usecnt > 0

- Properly log the app ports

	As result, some problems are fixed:

- possible use-after-free of svc in ip_vs_genl_set_cmd after
ip_vs_del_service because our usecnt reference does not guarantee that
svc is not freed on refcnt==0, eg. when no dests are moved to trash

- possible usecnt leak in do_ip_vs_set_ctl after ip_vs_del_service
when the service is not freed now, for example, when some
destionations are moved into trash and svc->refcnt remains above 0.
It is harmless because svc is not in hash anymore.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Patrick McHardy <kaber@trash.net>
2010-09-21 18:12:30 +02:00
Changli Gao
99f07e91be netfilter: save the hash of the tuple in the original direction for latter use
Since we don't change the tuple in the original direction, we can save it
in ct->tuplehash[IP_CT_DIR_REPLY].hnode.pprev for __nf_conntrack_confirm()
use.

__hash_conntrack() is split into two steps: hash_conntrack_raw() is used
to get the raw hash, and __hash_bucket() is used to get the bucket id.

In SYN-flood case, early_drop() doesn't need to recompute the hash again.

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
2010-09-21 17:49:20 +02:00
Takashi Iwai
abdd8f5106 ALSA: hda - Apply ALC269 VAIO fix-up to all Sony laptops with ALC269
We've applied a fix-up for ALC269 VAIO only for two models.  But all
Sony VAIO models with ALC269 codec seem to require the similar fix.
Let's apply it with vendor-id mask.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
2010-09-21 17:42:46 +02:00
Takashi Iwai
5e09416f36 Merge branch 'fix/hda' into topic/hda 2010-09-21 17:40:48 +02:00
Julian Anastasov
8a8030407f ipvs: make rerouting optional with snat_reroute
Add new sysctl flag "snat_reroute". Recent kernels use
ip_route_me_harder() to route LVS-NAT responses properly by
VIP when there are multiple paths to client. But setups
that do not have alternative default routes can skip this
routing lookup by using snat_reroute=0.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Patrick McHardy <kaber@trash.net>
2010-09-21 17:38:57 +02:00
Julian Anastasov
f4bc17cdd2 ipvs: netfilter connection tracking changes
Add more code to IPVS to work with Netfilter connection
tracking and fix some problems.

- Allow IPVS to be compiled without connection tracking as in
2.6.35 and before. This can avoid keeping conntracks for all
IPVS connections because this costs memory. ip_vs_ftp still
depends on connection tracking and NAT as implemented for 2.6.36.

- Add sysctl var "conntrack" to enable connection tracking for
all IPVS connections. For loaded IPVS directors it needs
tuning of nf_conntrack_max limit.

- Add IP_VS_CONN_F_NFCT connection flag to request the connection
to use connection tracking. This allows user space to provide this
flag, for example, in dest->conn_flags. This can be useful to
request connection tracking per real server instead of forcing it
for all connections with the "conntrack" sysctl. This flag is
set currently only by ip_vs_ftp and of course by "conntrack" sysctl.

- Add ip_vs_nfct.c file to hold all connection tracking code,
by this way main code should not depend of netfilter conntrack
support.

- Return back the ip_vs_post_routing handler as in 2.6.35 and use
skb->ipvs_property=1 to allow IPVS to work without connection
tracking

Connection tracking:

- most of the code is already in 2.6.36-rc

- alter conntrack reply tuple for LVS-NAT connections when first packet
from client is forwarded and conntrack state is NEW or RELATED.
Additionally, alter reply for RELATED connections from real server,
again for packet in original direction.

- add IP_VS_XMIT_TUNNEL to confirm conntrack (without altering
reply) for LVS-TUN early because we want to call nf_reset. It is
needed because we add IPIP header and the original conntrack
should be preserved, not destroyed. The transmitted IPIP packets
can reuse same conntrack, so we do not set skb->ipvs_property.

- try to destroy conntrack when the IPVS connection is destroyed.
It is not fatal if conntrack disappears before that, it depends
on the used timers.

Fix problems from long time:

- add skb->ip_summed = CHECKSUM_NONE for the LVS-TUN transmitters

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Patrick McHardy <kaber@trash.net>
2010-09-21 17:35:41 +02:00
Thomas Weber
6f0b31c318 Fix typo interrest[ing|ed] => interest[ing|ed]
Fix typos with interrest*.

Signed-off-by: Thomas Weber <weber@corscience.de>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2010-09-21 17:05:44 +02:00
Felix Fietkau
6e5c2b4e8a ath9k: make the driver specific rate control module optional
ath9k can use minstrel_ht instead, so it makes sense to save some space here.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:35 -04:00
Felix Fietkau
334b06029e ath9k: move ath_tx_aggr_check() to the rate control module
It is not used anywhere else and can be made static

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:34 -04:00
Felix Fietkau
90fa539ca3 ath9k: clean up / fix aggregation session flush
The tid aggregation cleanup is a bit fragile, as it discards failed
subframes in some places, and retransmits them in others. This could
block the cleanup of an existing aggregation session, if a retransmission
for a tid is issued, yet the tid is never scheduled again because of
the cleanup state.

Fix this by getting rid of as many subframes as possible, as early
as possible, and immediately transmitting pending subframes as regular
HT frames instead of waiting for the cleanup to complete.

Drop all pending subframes while keeping track of the Block ACK window
during aggregate tx completion to prevent sending out stale subframes,
which could confuse the receiver side.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:33 -04:00
Felix Fietkau
231c3a1f06 ath9k: fix an aggregation start related race condition
A new aggregation session start can be issued by mac80211, even when the
cleanup of the previous session has not completed yet. Since the data structure
for the session is not recreated, this could corrupt the block ack window
and lock up the aggregation session. Fix this by delaying the new session
until the old one has been cleaned up.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:31 -04:00
Felix Fietkau
81ee13ba7e ath9k: clean up block ack window handling
There's no reason to keep pointers to pending tx buffers around, if they're
only used to keep track of which frames are still pending. Use a bitfield
instead.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:30 -04:00
Eliad Peller
9ee82d5410 wl1271: bugfix: use bitwise-AND instead of logical-AND
typo - while looking for specific bits we should do a bitwise-AND instead of logical-AND.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Acked-by: Luciano Coelho <luciano.coelho@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:29 -04:00
Eliad Peller
58be4607d2 wl1271: avoid redundant memcpy of rx_status
copy the rx_status directly to skb->cb (control buffer) instead of copying
it to a local struct and then copying it again (for each rx packet)

Signed-off-by: Eliad Peller <eliad@wizery.com>
Acked-by: Luciano Coelho <luciano.coelho@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:28 -04:00
Christian Lamparter
2c7808d934 carl9170: fix hang in AP mode when HT STA does PSM
This is a obvious bug, skb_queue_walk does not
work if the iterator gets removed from the queue.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:26 -04:00
Christian Lamparter
735761108f carl9170: reinit phy after HT settings have changed
The driver has a set of different initvals for 20 MHz
vs dynamic HT2040 operation. Because we can't change
some of the registers "in-flight", the driver needs to
perform a warm reset.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:25 -04:00
Christian Lamparter
78ec789bd3 carl9170: fix state downgrade during reset
Don't mark the device as completely dead just yet.
If all goes to plan and carl9170_reboot succeeds
then we can skip the expensive userspace-driven
reinitialization anyway.

And if it doesn't and carl9170_reboot fails,
then carl9170_usb_cancel_urbs will do the
necessary steps.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:24 -04:00
Christian Lamparter
cf6487d0d6 carl9170: abort tasklet during usb reset
This patch prevents the tasklet code from
interfering while the firmware is down for
an unscheduled maintenance.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:23 -04:00
Christian Lamparter
9adc9e0ff3 carl9170: don't load bogus nf of chain 1
According to Atheros, chain 1 is not connected.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:21 -04:00
Christian Lamparter
e278c5a903 carl9170: fix noise dBm conversion
Ever since carl9170 gained support to read the noisefloor,
the reported noisefloor level was pretty poor.

Initially I assumed that something was wrong in the PHY
setup and it would be impossible to fix without any
guidances. But this was not the case. In fact the nf
readings were correct and the thing that was broken
was the "simple" sign extension code!

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:20 -04:00
Christian Lamparter
9dec6f9c48 carl9170: use rx chainmask from eeprom
The eeprom provides a mask for all present rx chains.
Why not use it instead of the generic initval default?

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:19 -04:00
Bruno Randolf
e0b1cc52e5 ath5k: Add tx queue configuration function
Add the mac80211 callback function to configure the tx queue properties like
cw_min, cw_max and aifs.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:18 -04:00
Bruno Randolf
de8af45520 ath5k: Simplify cw_min/max and AIFS configuration
Get rid of overly complicated cw_min/max and AIFS configuration:

* Validate values in ath5k_hw_set_tx_queueprops(), so we can use them directly
  without further checks or computation in ath5k_hw_reset_tx_queue().

* Simplifiy by using AR5K_TUNE_AIFS|CWMIN|CWMAX variables directly since we
  don't support XR or B channels. That way we can also remove
  AR5K_TXQ_USEDEFAULT and the confusing logic around it.

* Update data types: AIFS is u8, CW's are u16.

* Remove now unneeded variables in ath5k_hw.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:16 -04:00
Bruno Randolf
234132960d ath5k: Keep last descriptor in queue
If we return a TX descriptor to the pool of available descriptors, while a
queues TXDP still points to it we could potentially run into all sorts of
troube.

It has been suggested that there is hardware which can set the descriptors
done bit before it reads ds_link and moves on to the next descriptor. While the
documentation says this is not true for newer chipsets (the descriptor contents
are copied to some internal memory), we don't know about older hardware.

To be safe, we always keep the last descriptor in the queue, and avoid dangling
TXDP pointers. Unfortunately this does not fully resolve the problem - queues
still get stuck!

This is similar to what ath9k does.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:15 -04:00
Bruno Randolf
923e5b3d3d ath5k: Count how many times a queue got stuck
Add a counter to show how many times a queue got stuck in the debugfs queue
file.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:14 -04:00
Bruno Randolf
4edd761f40 ath5k: Add watchdog for stuck TX queues
Since we do not know any better solution to the problem that TX queues can get
stuck, this adds a timer-based watchdog, which will check for stuck queues and
reset the hardware if necessary.

Ported from ath9k commit 164ace3853.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:12 -04:00
Bruno Randolf
1440401e70 ath5k: Move tx frame completion into separate function
Clearer separation between queue handling and what we do with completed frames.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:11 -04:00
Bruno Randolf
651d9375dc ath5k: Fix TX queues stopping
It does not make sense to stop queues for NF calibration. This will not stop
transmissions from the card, if there are queued packets.

If we run out of TX buffers we need to stop all queues, not only one.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:10 -04:00
Bruno Randolf
cfddc11c42 ath5k: Fix queue debug file
Take txq lock in debug file and fix reporting of used buffers.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:09 -04:00
Bruno Randolf
925e0b0613 ath5k: Use four hardware queues
Prepare ath5k for WME by using four hardware queues.

The way we set up our queues matches the mac80211 queue priority 1:1, so we
don't have to do any mapping for queue numbers.

Every queue uses 50 of the total 200 available transmit buffers, so the DMA
memory usage does not increase with this patch, but it might be good to
fine-tune the number of buffers per queue later (depending on the CPU speed and
load, and the speed of the medium access, it might not be big enough).

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:08 -04:00
Bob Copeland
8a63facc37 ath5k: reorder base.c to remove fwd decls
This change reorganizes the main ath5k file in order to re-group
related functions and remove most of the forward declarations
(from 61 down to 3).  This is, unfortunately, a lot of churn, but
there should be no functional changes.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:06 -04:00
Bruno Randolf
ce2220d1da ath/ath5k/ath9k: Fix crypto capabilities merge issue
Fixing up a merge issue / concurrent development:

Remove unneeded ath_crypt_caps flags, as per "ath9k_hw: remove useless hw
capability flags" (364734fafb), but set the
AESCCM flag for ath9k. common ath code still needs a flag for this because
there is ath5k hardware which can't do AES in hardware.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:05 -04:00
Christian Lamparter
c807666a08 carl9170: update AR9170 phy initvals
This patch is loosely based on an ath9k patch called:
"ath9k_hw: sync initvals for ar9001 and ar9002 with Atheros"

It includes the following changes/fixes:
	- AGC setting improvements
	- timing changes for improved performance

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:04 -04:00
Ohad Ben-Cohen
80b517f362 omap: zoom: add mmc3/wl1271 device support
Add MMC3 support on ZOOM, which has the wl1271 device hardwired to.

The wl1271 is a 4-wire, 1.8V, embedded SDIO WLAN device with an
external IRQ line, and power-controlled by a GPIO-based fixed regulator.

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:02 -04:00
Ohad Ben-Cohen
b642fde7f1 omap: zoom: add fixed regulator device for wlan
Add a fixed regulator vmmc device to enable power control
of the wl1271 wlan device.

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:01 -04:00
Ohad Ben-Cohen
15cea99306 wl1271: make ref_clock configurable by board
The wl1271 device is using a reference clock that may change
between board to board.

Make the ref_clock parameter configurable by board settings
instead of having a hard coded value in the sources.

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:05:00 -04:00
Ohad Ben-Cohen
09cecc340b wl1271: take irq info from private board data
Remove the hard coded irq information, and instead take
the irq information from the board's platform data.

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:04:59 -04:00
Ohad Ben-Cohen
61ee7007a5 wl12xx: add platform data passing support
Add a simple mechanism to pass platform data to the
SDIO instances of wl12xx.

This way there is no confusion over who owns the 'embedded data',
typechecking is preserved, and no possibility for the wrong driver to
pick up the data.

Originally proposed by Russell King.

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:04:57 -04:00
Ohad Ben-Cohen
2cc78ff78c wl1271: propagate set_power's return value
Make it possible for the set power method to indicate a
success/failure return value. This is needed to support
more complex power on/off operations such as SDIO
power manipulations.

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Acked-by: Luciano Coelho <luciano.coelho@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:04:56 -04:00
Nikanth Karthikesan
817f2c842d Fix various typos of valid in comments
Fix various typos of valid.

Signed-off-by: Nikanth Karthikesan <knikanth@suse.de>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2010-09-21 17:04:50 +02:00
Ohad Ben-Cohen
c1f9a09560 wl12xx: make wl12xx.h common to both spi and sdio
Move wl12xx.h outside of the spi-specific location,
so it can be shared with both spi and sdio solutions.

Update all users of spi/wl12xx.h accordingly

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Acked-by: Luciano Coelho <luciano.coelho@nokia.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-09-21 11:04:38 -04:00
Nikanth Karthikesan
0bc79f7f58 Doc: Fix typo s/packages/packaged
Fix typo s/packages/packaged in Documentation/vm/numa_memory_policy.txt.

Signed-off-by: Nikanth Karthikesan <knikanth@suse.de>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2010-09-21 17:03:27 +02:00
Steven Rostedt
a8027073eb tracing/sched: Add sched_pi_setprio tracepoint
Add a tracepoint that shows the priority of a task being boosted
via priority inheritance.

Cc: Gregory Haskins <ghaskins@novell.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2010-09-21 10:56:41 -04:00
Henrik Rydberg
46c4ba012c HID: 3m: Output proper orientation range
The range of orientation values for height/width devices should
be [0, 1], but is currently set to [1, 1]. Having min == max also
breaks uinput device setup. Fixed with this patch.

Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2010-09-21 16:16:09 +02:00
Henrik Rydberg
41035901df HID: 3m: Adjust to sequential MT HID protocol
The multitouch extensions to the HID protocol allows for contact
data to be sent over several reports, which is also the case for
the 3M M2256PW touchscreen. This patch modifies the logic to only
synchronize the input layer when all contacts have been received.
Consequentially, the full 60-finger capacity of the device is enabled.

Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
Acked-by: Stephane Chatty <chatty@enac.fr>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2010-09-21 16:11:44 +02:00
Henrik Rydberg
24750f3e46 HID: Add a hid quirk for input sync override
As of lately, HID devices which send per-frame data split over several
HID reports have started to emerge. This patch adds a quirk which
allows the HID driver to take over the input layer synchronization,
and hence the control of the frame boundary.

Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2010-09-21 16:11:44 +02:00
Chris Wilson
9e0ae53404 drm/i915: Don't overwrite the returned error-code
During i915_gem_create_mmap_offset() if the subsystem reports an error
code, use it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
2010-09-21 15:05:24 +01:00
Chris Wilson
4fd21dc8ee drm/i915/lvds: Unlock the PP register when panel-fitting
As we do not wait for the panel to turn off when we need to adjust the
panel-fitting registers we also need to unlock the PLLs as with the
non-pfit update path.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
2010-09-21 14:22:54 +01:00
Steven Rostedt
b3bc211cfe sched: Give CPU bound RT tasks preference
If a high priority task is waking up on a CPU that is running a
lower priority task that is bound to a CPU, see if we can move the
high RT task to another CPU first. Note, if all other CPUs are
running higher priority tasks than the CPU bounded current task,
then it will be preempted regardless.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Gregory Haskins <ghaskins@novell.com>
LKML-Reference: <20100921024138.888922071@goodmis.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-09-21 13:57:12 +02:00
Steven Rostedt
43fa5460fe sched: Try not to migrate higher priority RT tasks
When first working on the RT scheduler design, we concentrated on
keeping all CPUs running RT tasks instead of having multiple RT
tasks on a single CPU waiting for the migration thread to move
them. Instead we take a more proactive stance and push or pull RT
tasks from one CPU to another on wakeup or scheduling.

When an RT task wakes up on a CPU that is running another RT task,
instead of preempting it and killing the cache of the running RT
task, we look to see if we can migrate the RT task that is waking
up, even if the RT task waking up is of higher priority.

This may sound a bit odd, but RT tasks should be limited in
migration by the user anyway. But in practice, people do not do
this, which causes high prio RT tasks to bounce around the CPUs.
This becomes even worse when we have priority inheritance, because
a high prio task can block on a lower prio task and boost its
priority. When the lower prio task wakes up the high prio task, if
it happens to be on the same CPU it will migrate off of it.

But in reality, the above does not happen much either, because the
wake up of the lower prio task, which has already been boosted, if
it was on the same CPU as the higher prio task, it would then
migrate off of it. But anyway, we do not want to migrate them
either.

To examine the scheduling, I created a test program and examined it
under kernelshark. The test program created CPU * 2 threads, where
each thread had a different priority. The program takes different
options. The options used in this change log was to have priority
inheritance mutexes or not.

All threads did the following loop:

static void grab_lock(long id, int iter, int l)
{
	ftrace_write("thread %ld iter %d, taking lock %d\n",
		     id, iter, l);
	pthread_mutex_lock(&locks[l]);
	ftrace_write("thread %ld iter %d, took lock %d\n",
		     id, iter, l);
	busy_loop(nr_tasks - id);
	ftrace_write("thread %ld iter %d, unlock lock %d\n",
		     id, iter, l);
	pthread_mutex_unlock(&locks[l]);
}

void *start_task(void *id)
{
	[...]
	while (!done) {
		for (l = 0; l < nr_locks; l++) {
			grab_lock(id, i, l);
			ftrace_write("thread %ld iter %d sleeping\n",
				     id, i);
			ms_sleep(id);
		}
		i++;
	}
	[...]
}

The busy_loop(ms) keeps the CPU spinning for ms milliseconds. The
ms_sleep(ms) sleeps for ms milliseconds. The ftrace_write() writes
to the ftrace buffer to help analyze via ftrace.

The higher the id, the higher the prio, the shorter it does the
busy loop, but the longer it spins. This is usually the case with
RT tasks, the lower priority tasks usually run longer than higher
priority tasks.

At the end of the test, it records the number of loops each thread
took, as well as the number of voluntary preemptions, non-voluntary
preemptions, and number of migrations each thread took, taking the
information from /proc/$$/sched and /proc/$$/status.

Running this on a 4 CPU processor, the results without changes to
the kernel looked like this:

Task        vol    nonvol   migrated     iterations
----        ---    ------   --------     ----------
  0:         53      3220       1470             98
  1:        562       773        724             98
  2:        752       933       1375             98
  3:        749        39        697             98
  4:        758         5        515             98
  5:        764         2        679             99
  6:        761         2        535             99
  7:        757         3        346             99

total:     5156       4977      6341            787

Each thread regardless of priority migrated a few hundred times.
The higher priority tasks, were a little better but still took
quite an impact.

By letting higher priority tasks bump the lower prio task from the
CPU, things changed a bit:

Task        vol    nonvol   migrated     iterations
----        ---    ------   --------     ----------
  0:         37      2835       1937             98
  1:        666      1821       1865             98
  2:        654      1003       1385             98
  3:        664       635        973             99
  4:        698       197        352             99
  5:        703       101        159             99
  6:        708         1         75             99
  7:        713         1          2             99

total:     4843       6594      6748            789

The total # of migrations did not change (several runs showed the
difference all within the noise). But we now see a dramatic
improvement to the higher priority tasks. (kernelshark showed that
the watchdog timer bumped the highest priority task to give it the
2 count. This was actually consistent with every run).

Notice that the # of iterations did not change either.

The above was with priority inheritance mutexes. That is, when the
higher prority task blocked on a lower priority task, the lower
priority task would inherit the higher priority task (which shows
why task 6 was bumped so many times). When not using priority
inheritance mutexes, the current kernel shows this:

Task        vol    nonvol   migrated     iterations
----        ---    ------   --------     ----------
  0:         56      3101       1892             95
  1:        594       713        937             95
  2:        625       188        618             95
  3:        628         4        491             96
  4:        640         7        468             96
  5:        631         2        501             96
  6:        641         1        466             96
  7:        643         2        497             96

total:     4458       4018      5870            765

Not much changed with or without priority inheritance mutexes. But
if we let the high priority task bump lower priority tasks on
wakeup we see:

Task        vol    nonvol   migrated     iterations
----        ---    ------   --------     ----------
  0:        115      3439       2782             98
  1:        633      1354       1583             99
  2:        652       919       1218             99
  3:        645       713        934             99
  4:        690         3          3             99
  5:        694         1          4             99
  6:        720         3          4             99
  7:        747         0          1            100

Which shows a even bigger change. The big difference between task 3
and task 4 is because we have only 4 CPUs on the machine, causing
the 4 highest prio tasks to always have preference.

Although I did not measure cache misses, and I'm sure there would
be little to measure since the test was not data intensive, I could
imagine large improvements for higher priority tasks when dealing
with lower priority tasks. Thus, I'm satisfied with making the
change and agreeing with what Gregory Haskins argued a few years
ago when we first had this discussion.

One final note. All tasks in the above tests were RT tasks. Any RT
task will always preempt a non RT task that is running on the CPU
the RT task wants to run on.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Gregory Haskins <ghaskins@novell.com>
LKML-Reference: <20100921024138.605460343@goodmis.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-09-21 13:57:12 +02:00