Glauber Costa [Sun, 11 Dec 2011 21:47:02 +0000 (21:47 +0000)]
foundations of per-cgroup memory pressure controlling.
This patch replaces all uses of struct sock fields' memory_pressure,
memory_allocated, sockets_allocated, and sysctl_mem to acessor
macros. Those macros can either receive a socket argument, or a mem_cgroup
argument, depending on the context they live in.
Since we're only doing a macro wrapping here, no performance impact at all is
expected in the case where we don't have cgroups disabled.
Laszlo Ersek [Sun, 11 Dec 2011 01:48:59 +0000 (01:48 +0000)]
xen-netfront: delay gARP until backend switches to Connected
After a guest is live migrated, the xen-netfront driver emits a gratuitous
ARP message, so that networking hardware on the target host's subnet can
take notice, and public routing to the guest is re-established. However,
if the packet appears on the backend interface before the backend is added
to the target host's bridge, the packet is lost, and the migrated guest's
peers become unable to talk to the guest.
A sufficient two-parts condition to prevent the above is:
(1) ensure that the backend only moves to Connected xenbus state after its
hotplug scripts completed, ie. the netback interface got added to the
bridge; and
(2) ensure the frontend only queues the gARP when it sees the backend move
to Connected.
These two together provide complete ordering. Sub-condition (1) is already
satisfied by commit f942dc2552b8 in Linus' tree, based on commit 6b0b80ca7165 from [1].
In general, the full condition is sufficient, not necessary, because,
according to [2], live migration has been working for a long time without
satisfying sub-condition (2). However, after 6b0b80ca7165 was backported
to the RHEL-5 host to ensure (1), (2) still proved necessary in the RHEL-6
guest. This patch intends to provide (2) for upstream.
It looks like the regression was introduced between 20111202 and 20111205 (linux-next tree). Symptoms: connection to AP seem to be
established, but no data goes though it in any way. Tested on intel
5300.
Peek at the changes have shown that it looks like at least part of
the code wasn't merged properly. It was originally committed into
iwl_agn.c but code in question was moved to iwl-mac80211.c.
This patch puts code in place and my card works again.
wl12xx: silence tx_attr uninitialized warning in wl1271_tx_fill_hdr
CC [M] drivers/net/wireless/wl12xx/tx.o
drivers/net/wireless/wl12xx/tx.c: In function ‘wl1271_tx_fill_hdr’:
drivers/net/wireless/wl12xx/tx.c:288:6: warning: ‘tx_attr’ may be used uninitialized in this function
Pavel Emelyanov [Fri, 9 Dec 2011 06:24:06 +0000 (06:24 +0000)]
udp_diag: Implement the get_exact dumping functionality
Do the same as TCP does -- lookup a socket in the given udp_table,
check cookie, fill the reply message with existing inet socket dumping
helper and send one back.
Pavel Emelyanov [Fri, 9 Dec 2011 06:23:00 +0000 (06:23 +0000)]
inet_diag: Introduce the inet socket dumping routine
The existing inet_csk_diag_fill dumps the inet connection sock info
into the netlink inet_diag_message. Prepare this routine to be able
to dump only the inet_sock part of a socket if the icsk part is missing.
This will be used by UDP diag module when dumping UDP sockets.
Pavel Emelyanov [Fri, 9 Dec 2011 06:22:26 +0000 (06:22 +0000)]
inet_diag: Split inet_diag_get_exact into parts
Similar to previous patch: the 1st part locks the inet handler
and will get generalized and the 2nd one dumps icsk-s and will
be used by TCP and DCCP handlers.
Pavel Emelyanov [Fri, 9 Dec 2011 06:22:10 +0000 (06:22 +0000)]
inet_diag: Split inet_diag_get_exact into parts
The 1st part locks the inet handler and the 2nd one dump the
inet connection sock.
In the next patches the 1st part will be generalized to call
the socket dumping routine indirectly (i.e. TCP/UDP/DCCP) and
the 2nd part will be used by TCP and DCCP handlers.
Pavel Emelyanov [Fri, 9 Dec 2011 06:21:16 +0000 (06:21 +0000)]
inet_diag: Remove indirect sizeof from inet diag handlers
There's an info_size value stored on inet_diag_handler, but for existing
code this value is effectively constant, so just use sizeof(struct tcp_info)
where required.
This commit breaks the build with CONFIG_NETPRIO_CGROUP=y so
revert it. It does build as a module though. The SUBSYS macro
in the cgroup core code automatically defines a subsys structure
as extern. Long term we should fix the macro. And I need to
fully build test things.
Tested with CONFIG_NETPRIO_CGROUP={y|m|n} with and without
CONFIG_CGROUPS defined.
Dmitry Kravkov [Thu, 8 Dec 2011 08:04:07 +0000 (08:04 +0000)]
bnx2x: properly initialize L5 features
The code is missing initialization of NO_FCOE_FLAG and NO_ISCSI*FLAGS
when CONFIG_CNIC is not selected.
This causes panic during driver load since commit 1d187b34daaecbb87aa523ba46b92930a388cb21 where NO_FCOE tested
unconditionally (outside #ifdef BCM_CNIC structure) and
accessed fp[FCOE_IDX] which is not allocated.
Eric Dumazet [Thu, 8 Dec 2011 06:06:03 +0000 (06:06 +0000)]
sch_red: Adaptative RED AQM
Adaptative RED AQM for linux, based on paper from Sally FLoyd,
Ramakrishna Gummadi, and Scott Shenker, August 2001 :
http://icir.org/floyd/papers/adaptiveRed.pdf
Goal of Adaptative RED is to make max_p a dynamic value between 1% and
50% to reach the target average queue : (max_th - min_th) / 2
Every 500 ms:
if (avg > target and max_p <= 0.5)
increase max_p : max_p += alpha;
else if (avg < target and max_p >= 0.01)
decrease max_p : max_p *= beta;
target :[min_th + 0.4*(min_th - max_th),
min_th + 0.6*(min_th - max_th)].
alpha : min(0.01, max_p / 4)
beta : 0.9
max_P is a Q0.32 fixed point number (unsigned, with 32 bits mantissa)
Changes against our RED implementation are :
max_p is no longer a negative power of two (1/(2^Plog)), but a Q0.32
fixed point number, to allow full range described in Adatative paper.
To deliver a random number, we now use a reciprocal divide (thats really
a multiply), but this operation is done once per marked/droped packet
when in RED_BETWEEN_TRESH window, so added cost (compared to previous
AND operation) is near zero.
dump operation gives current max_p value in a new TCA_RED_MAX_P
attribute.
Example on a 10Mbit link :
tc qdisc add dev $DEV parent 1:1 handle 10: est 1sec 8sec red \
limit 400000 min 30000 max 90000 avpkt 1000 \
burst 55 ecn adaptative bandwidth 10Mbit
# tc -s -d qdisc show dev eth3
...
qdisc red 10: parent 1:1 limit 400000b min 30000b max 90000b ecn
adaptative ewma 5 max_p=0.113335 Scell_log 15
Sent 50414282 bytes 34504 pkt (dropped 35, overlimits 1392 requeues 0)
rate 9749Kbit 831pps backlog 72056b 16p requeues 0
marked 1357 early 35 pdrop 0 other 0
Jiri Pirko [Thu, 8 Dec 2011 04:11:20 +0000 (04:11 +0000)]
team: use vlan_vids_[addr/del]_by_dev
So far when vlan id was added to team device befor port was added, this
vid was not added to port's vlan filter. Also after removal, vid stayed
in port device's vlan filter. Benefit of new vlan functions to handle
this work.
Jiri Pirko [Fri, 9 Dec 2011 00:52:37 +0000 (19:52 -0500)]
net: make vlan ndo_vlan_rx_[add/kill]_vid return error value
Let caller know the result of adding/removing vlan id to/from vlan
filter.
In some drivers I make those functions to just return 0. But in those
where there is able to see if hw setup went correctly, return value is
set appropriately.
Lothar Waßmann [Wed, 7 Dec 2011 21:59:31 +0000 (21:59 +0000)]
net/fec: fix the .remove code
The .remove code is broken in several ways.
- mdiobus_unregister() is called twice for the same object in case of dual FEC
- phy_disconnect() is being called when the PHY is already disconnected
- the requested IRQ(s) are not freed
- fec_stop() is being called with the inteface already stopped
All of those lead to kernel crashes if the remove function is actually used.
Lothar Waßmann [Wed, 7 Dec 2011 21:59:29 +0000 (21:59 +0000)]
net/fec: don't munge MAC address from platform data
When the MAC address is supplied via platform_data it should be OK as
it is and should not be modified in case of a dual FEC setup.
Also copying the MAC from platform_data to the single 'macaddr'
variable will overwrite the MAC for the first interface in case of a
dual FEC setup.
Matt Carlson [Thu, 8 Dec 2011 14:40:14 +0000 (14:40 +0000)]
tg3: Save stats across chip resets
Tg3 has a place to store stats, but doesn't really use it. This patch
modifies the driver so that stats are saved across chip resets and gets
cleared across close / open calls.
bridge: add local MAC address to forwarding table (v2)
If user has configured a MAC address that is not one of the existing
ports of the bridge, then we need to add a special entry in the forwarding
table. This forwarding table entry has no outgoing port so it has to be
treated a little differently. The special entry is reported by the netlink
interface with ifindex of bridge, but ignored by the old interface since there
is no usable way to put it in the ABI.
Pass bridge to fdb_notify so it can determine correct namespace based
on namespace of bridge rather than namespace of destination port.
Also makes next patch easier.
Eyal Shapira [Wed, 7 Dec 2011 10:37:04 +0000 (12:37 +0200)]
wl12xx: minor fix in sched_scan_ssid_list
The user can pass broadcast SSID (ssid="") in the list of SSIDs for active scan.
In this case the loop was attempting to match SSIDs in the filter
list to this empty entry and marking them as HIDDEN (sending probe
request) by mistake
Eliad Peller [Tue, 6 Dec 2011 10:15:06 +0000 (12:15 +0200)]
wl12xx: remove redundant commands from plt init
During plt init we configure some redundant commands,
which are not needed for plt (specifically, we shouldn't
configure any role-specific params, as there are no
active roles). remove them.
Eliad Peller [Tue, 6 Dec 2011 10:15:05 +0000 (12:15 +0200)]
wl12xx: fix testmode test/interrogate commands
fix several issues in testmode test/interrogate commands:
1. check the driver state is not OFF.
2. wakeup the chip from elp (if needed)
3. fix memory leak in wl1271_tm_cmd_interrogate()
Enable ANI for ar9100 since it seems to be working fine (and as a
matter of fact ANI was always performed for ar9100 since code which
was supposed to disable it didn't achieve this goal).
This patch sets config.enable_ani to default (true) value for
ar9100.
ath9k: use config.enable_ani to check if ani should be performed
Currently in ath9k code there is an attempt which is meant to
disable ANI for ar9100 and ar9340. But it doesn't really achieve
this. All it does is disable ANI init and setup (i.e. calls to
ath9k_hw_ani_setup and ath9k_hw_ani_init). Since ath9k_hw_ani_setup is
not called ah->config.ani_poll_interval is never initialized (i.e. it
is always zero) and ath_ani_calibrate always executes ANI procedures
(over uninitialized ANI parameters).
Moreover, ath_ani_calibrate is being called each 1ms because
common->ani.timer is set to zero interval because
ah->config.ani_poll_interval==0 (and thus smallest value of all
intervals). Normally it should not be called this often.
This patch changes the code so config.enable_ani is used to check if
ANI should be performed.
config.enable_ani is initialized to true by default. This patch sets
it to false for ar9100 and ar9340.
Felix Fietkau [Wed, 16 Nov 2011 12:08:42 +0000 (13:08 +0100)]
ath9k: only drop packets in drv_flush when asked to
Recently more places in mac80211 call drv_flush to ensure proper order
for state changes wrt. powersave, channel changes, etc. On some systems
such calls lead to spurious logspam about failing to stop tx dma, as well
as hardware resets that go along with that.
Instead of dropping packets in a place where it's completely unnecessary,
only do it when drop == true.
Felix Fietkau [Wed, 16 Nov 2011 12:08:41 +0000 (13:08 +0100)]
ath9k: rework power state handling
Turning off the radio when mac80211 tells the driver that it's idle is not
a good idea, as idle interfaces might still occasionally scan or send packets.
The only time the radio can be safely turned off is when drv_stop has been
called. In the mean time, use sc->ps_idle only to indicate network sleep vs
full sleep.
Move the LED GPIO changes out of the PCI suspend/resume path, the start/stop
functions already take care of that.
Felix Fietkau [Wed, 16 Nov 2011 12:08:40 +0000 (13:08 +0100)]
ath9k: always issue a full hw reset after waking up from full-sleep mode
After waking up from full sleep, registers are accessible, but rx/tx
typically fails. A fast channel change will not recover from this, so
ensure that a full-sleep -> wake transition is always followed by a full
reset.
The reason why this hasn't created any serious problems yet is that it's
hidden by the (wrong) behavior of enabling/disabling the radio when the
wiphy idle state changes.
Various bugfixes for caif_shmcore.c:
- fix deadlocks due to improper usage of spin-lock
- add missing spin-lock init
- don't call dev_kfree_skb() with irqs disabled,
use dev_kfree_skb_irq() instead.
- fix potential skb null pointer de-reference.
Christian Auby [Tue, 6 Dec 2011 12:15:41 +0000 (12:15 +0000)]
caif-hsi: Remove wake line modification when flushing FIFO
Raising wake before flushing FIFO and lowering it after caused a
spike on AC wake that were sometimes detected and acted upon by the
modem. Fixed this by remove wake line modification when flushing FIFO.
Michael Chan [Tue, 6 Dec 2011 10:58:08 +0000 (10:58 +0000)]
bnx2x: Fix compile errors if CONFIG_CNIC is not set
Don't provide FCoE and iSCSI statistics to management firmware if
CONFIG_CNIC is not set. Some needed structure fields are not defined
without CONFIG_CNIC.
Johannes Berg [Tue, 6 Dec 2011 09:39:40 +0000 (10:39 +0100)]
mac80211: accept public action frames with mismatched BSSID
Arik's patch "mac80211: allow action frames with unknown
BSSID in GO mode" allowed any action frames in P2P mode
to go through, but only to cooked monitor interfaces as
the IEEE80211_RX_RA_MATCH was still cleared. As a result
my no-monitor patches broke invitation responses.
Instead of allowing any action frames in P2P GO mode to
go through with a wrong BSSID like that patch did, allow
all public action frames. They will never be processed
by mac80211, but can be reported via nl80211 then.
Johannes Berg [Fri, 2 Dec 2011 21:08:52 +0000 (22:08 +0100)]
mac80211: fix retransmit
This fixes another regression from my "pass all
fragments to driver at once" patches -- if the
packet is being retransmitted then we don't go
through all handlers, but we still need to move
it to the skbs list, otherwise we run into the
first warning in __ieee80211_tx() and leak the
skb.
Don Fry [Fri, 2 Dec 2011 16:48:38 +0000 (08:48 -0800)]
iwlwifi: replace iwl_priv reference with iwl_trans for ucode.
Replace the references to the iwl_priv structure with the iwl_trans
structure as the priv structure is never referenced other than to
access the trans structure. Rename from iwlagn to iwl.
Kenny Hsu [Fri, 2 Dec 2011 16:48:27 +0000 (08:48 -0800)]
iwlwifi: add range checking in tm sram read command
The size of sram may alter according to ucode type.
Retrieve the maximum sram size by current ucode
type for range checking to prevent wrong data access.
Hsu, Kenny [Fri, 2 Dec 2011 16:48:26 +0000 (08:48 -0800)]
iwlwifi: add tm commands for sram reading by dumpit
Create new testmode commands and attributes to suppot
sram data reading. Because the amount of sram data
may exceed single skb packet size. Using the nl80211
dump it funtion to deliver sram data to userspace.
based on systems change to improve rx dynamic range,
and enables heavy clip for 5G HT40 MCS0 to improve
spectral mask power. also remove an unused function
declaration
Alexander Simon [Wed, 30 Nov 2011 15:56:34 +0000 (16:56 +0100)]
mac80211: Add HT operation modes for IBSS
The HT mode is set by iw (previous patchsets).
The interface is set into the specified HT mode.
HT mode and capabilities are announced in beacons.
If we add a station that uses HT also, the fastest matching HT mode will
be used for transmission. That means if we are using HT40+ and we add a station
running on HT40-, we would transfer at HT20.
If we join an IBSS with HT40, but the secondary channel is not
available, we will fall back into HT20 as well.
Allow frame aggregation to start in IBSS mode.
Signed-off-by: Alexander Simon <[email protected]>
[[email protected]: Updates]
* remove implicit channel_type enum assumptions
* use rate_control_rate_init() if channel type changed
* remove channel flags check
* activate HT IBSS feature support
* slightly reword commit message
* rebase on wireless-testing
projects / linux.git / log