I have seen RX stalls on a machine that experienced a suspected
OOM. After the stall, the RX buffer is empty on the guest side
and there are exactly 16 entries available on the host side. As
the number of entries is less than that required by a maximal
skb, the host cannot proceed.
The guest did not have a refill job scheduled.
My diagnosis is that an OOM had occured, with the delayed refill
job scheduled. The job was able to allocate at least one skb, but
not enough to overcome the minimum required by the host to proceed.
As the refill job would only reschedule itself if it failed completely
to allocate any skbs, this would lead to an RX stall.
The following patch removes this stall possibility by always
rescheduling the refill job until the ring is totally refilled.
Testing has shown that the RX stall no longer occurs whereas
previously it would occur within a day.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
The low-level MCDI code always uses 32-bit MMIO operations, and
callers must pad input and output buffers to multiples of 4 bytes.
The MCDI NVRAM functions are not doing this. Also, their buffers are
declared as variable-length arrays with no explicit maximum length.
Switch to a fixed buffer size based on the chunk size used by the
MTD driver (which is a multiple of 4).
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to a hardware bug in the SFC9000 family, the firmware must
transfer raw GMAC statistics to host memory before aggregating them
into the cooked (speed-independent) MAC statistics. Extend the stats
buffer to support this.
The length of the buffer is explicit in the MAC_STATS command, so this
change is backward-compatible on both sides.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
By rounding up the buffer size to power of 2, several expensive
modulus operations can be avoided. This patch also solves a bug where
the gap need when ring gets full was not being accounted for.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently qlge tries to release regions even if they were not allocated.
This causes messages like the following in the kernel log
Trying to free nonexistent resource <00000000006af400-00000000006af4ff>
Trying to free nonexistent resource <00003c04ff9f4000-00003c04ff9f7fff>
Trying to free nonexistent resource <00003c04ffc00000-00003c04ffcfffff>
This patch fixes the goto logic in order to not release the resources
if they were not allocated.
Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Filtered frames not only need their control information
cleared to avoid wrong checks, but also need to have
jiffies and vif assigned so they can be processed or
expired.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Currently, the remain_on_channel work callback needs
to track in its own data structure whether the work
was just started or not. By reordering some code this
becomes unnecessary, the generic wk->started variable
can still be 'false' on the first invocation and only
be 'true' on actual timeout invocations, so that the
extra variable can be removed.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
What a stupid mistake. In
commit 813d766940
Author: Johannes Berg <johannes@sipsolutions.net>
Date: Sun Jan 17 01:47:58 2010 +0100
mac80211: move control.hw_key assignment
I inserted code testing the wrong flags field,
which means that the test is almost always true
(it's really testing for the peer's WMM support)
and thus the later parts of the stack assume hw
crypto will be done even if that's not true.
Obviously, that broke software crypto. Maxim
said so specifically, and Jochen probably uses
some cipher that iwl3945 doesn't support in
hardware, which might also explain that Maxim
reports that even hw crypto is broken.
Fix this to test the right flags field.
Reported-by: Maxim Levitsky <maximlevitsky@gmail.com>
Reported-by: Jochen Friedrich <jochen@scram.de>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
ieee80211_work_rx_mgmt currently enqueues various management frames,
including deauth and disassoc frames, however the function
ieee80211_work_rx_queued_mgmt does not handle these, as they should
only occur if the AP is buggy. It does emit a WARN_ON when this happens
though, and several users have reported such instances.
Fix the WARN_ON by not queueing such frames in the first place.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Johannes noticed that I had incorrectly documented the context of
update_tkip_key() driver operation. It must be atomic because all
RX code is run inside rcu critical section.
Reported-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Kalle Valo <kalle.valo@iki.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The TKIP key update callback is called from the RX path, where the driver
mutex is already locked. This results in a circular locking bug.
Avoid this by removing the lock.
Johannes noted that there is a separate bug: The callback still breaks on SDIO
hardware, because SDIO hardware access needs to sleep, but we are not allowed
to sleep in the callback due to mac80211's RCU locking.
Signed-off-by: Michael Buesch <mb@bu3sch.de>
Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Reported-by: kecsa@kutfo.hit.bme.hu
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: stable <stable@kernel.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
ath9k currently forces hw->max_rate_tries to 4 to work around rate
control inefficiencies. This has some negative side effects, such as
rate_control_send_low also using a maximum of 4 tries, which could
negatively affect reliability of unicast management frames.
This patch pushes the retry limit to the rate control instead, and
allows it to use more tries on the last stage to prevent unnecessary
packet loss.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The following commit
commit e4da8c37af
Author: Johannes Berg <johannes@sipsolutions.net>
Date: Wed Dec 23 13:15:43 2009 +0100
mac80211: make off-channel work generic
triggered a bug in iwlwifi where HT parameters would not be correctly set in
some mac80211 pathways. The aggregation (and possibly other) station flags were
not being set, which limited the size of aggregation blocks and reduced
throughput at high rates.
>From Johannes:
"""
Due to Wey-Yi's patch to use the set-channel command when the channel changes
while associated, we don't get a full new RXON. Therefore, we don't re-set the
rxon-station either. However, under some circumstances that apparently have
gotten more likely mac80211 will first set up the BSS info, then add the
station and then switch to an HT channel type.
Therefore, the check for "priv->current_ht_config.is_ht" in
iwl_rxon_add_station() will hit false and not fill in the HT information.
However, that check can just be removed, which is the easiest fix for all this,
because the HT capa struct is always there, just could possibly have the
ht_supported member set to false.
"""
A sample good link in my 3x3 network improves by approximately 25% TCP
throughput. This fixes Bug 2144
(http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2144).
Signed-off-by: Daniel Halperin <dhalperi@cs.washington.edu>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
In iwlagn the support for spectrum measurement command has been
disabled since v2.6.29 without any requests for it. In addition to this
when this command is indeed enabled it has been found to trigger firmware
SYSASSERT on at least 4965 and 5100 hardware (see
http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=1952 ). Since then
this code has been bitrotting and cannot just be enabled without porting.
Remove support for spectrum measurement command from iwlagn. It can be
added back if there is a future need and the firmware problem it triggers
has been fixed. Support for the spectrim measurement notification remains
as it has been enabled all the time.
In addition to this remove the 3945 spectrum measurement command Kconfig
option and make this command always supported. The code added by this
enabling is minimal and only run when user triggers a spectrum measurement
request via sysfs.
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
sparse won't check endianness annotations by
default, but iwlwifi is and should be clean
so we can make sparse check them on it.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
sparse correctly warns about symbol not
being static, make static to shut it up.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The debugfs code can be made a whole lot more
efficient by using debugfs_remove_recursive(),
the large chunk of variables can completely go
away and by moving two variables we no longer
need to allocate an extra chunk of memory.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
It is better to first notify cfg80211 about the hw
rfkill state (so the rfkill device that will be
registered won't have the wrong state while being
registered), and the power/tt variable init can
(and probably should) also be done first.
Also rename iwl_setup_mac to
iwl_mac_setup_register to better describe what it
really does.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
It really hasn't been for a long time, not sure
why this stuck around.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Adding "bt_coex_active" module parameter for iwlcore to enable/disable
BT coexist; if bt_coex_active is true (default), uCode will do kill/defer
every time the priority line is asserted (BT is sending signals on the
priority line in the PCIx). By disable the bt_coex_active, uCode will
ignore the BT activity and perform the normal operation.
Users might experience transmit issue on some platform due to this
WiFi/BT co-exist problem. The possible symptoms are: NetworkManager and
other similar programs can scan and find all the available APs, but will
timeout and unable to associate with any of the APs; no out-going frames
can be found with wireless sniffer tools.
On those platforms, WiFi communication can be restored by set "bt_coex_active"
module parameter to "false"
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Give better bit filed define and description for flag parameter
in REPLY_BT_CONFIG command:
flags:
bit 0 - 1: BT channel announcement enabled
0: disable
bit 1 - 1: priority of BT device enabled
0: disable
bit 2 - 1: BT 2 wire support enabled
0: disable
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add function pointer for broadcast station addition so that we can call it
in from iwlcore at a later time. We only distinguish between iwlagn and
iwl3945 broadcast station addition. For the iwl3945 station addition we add
that function to iwlcore since that is where most station functionality
resides, making it part of iwl3945 will require significant code
reorganization that will dilute station management functionality. This
seems to be an efficient solution.
It may seem as though we are removing error checking when adding the 3945
broadcast station but this error checking was never really necessary since
the function returns the station id and the broadcast station id is always
set.
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The work done when a station is added is very similar whether the station
is added synchronously or asynchronously. Centralize this work. At the same
time increase the status flags being checked for when the command returns
with accompanying debug messages. Also increase checking when setting the
"ucode active" state with accompanying debugging.
This work is done in preparation for station notification support.
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
From: Abhijeet Kolekar <abhijeet.kolekar@intel.com>
For 6X50 DC calibration needs to be initialized
else uCode will run an endless loop.
Enbale DC calibration in hw config.
Signed-off-by: Abhijeet Kolekar <abhijeet.kolekar@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
We have seen the throughput dropped due to external noisy environment
and the radio is out of tune. There are lot of plcp errors indicating
this condition. Eventually the station can get de-authenticated by the
Access Point. By resetting and tuning the radio, the plcp errors are
reduced or eliminated and the throughput starts to rise.
To prevent unexpected behavior such as drop in throughput or deauthentication,
- The change provides the driver feature to monitor and tune the radio base on
the statistics notification from the uCode.
- It also allows the setting of the plcp error rate threshold via
the plcp_delta under debugfs interface.
Signed-off-by: Trieu 'Andrew' Nguyen <trieux.t.nguyen@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add additional logic for internal scan routine to control how
frequent this function should be performed.
The intent of this function is to reset/re-tune the radio and bring the
RF/PHY back to normal state, it does not make sense calling it too
frequent,
if reset the radio can not bring it back to normal state, it indicate
there are other reason to cause the radio not operate correctly.
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch allows A band to be scanned when driver is associated to AP.
Scan mechanism is that mac80211/cfg80211 requests driver to scan G band
first and then immediately to scan A band. Original code require
driver to wait for 2 seconds after any scan before another scan will be
performed. This caused driver to service G band scan request from
mac80211/cfg80211 but deny the A band scan request.
Signed-off-by: Henry Zhangh <hongx.c.zhang@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Adding "radio reset" function to help reset and stabilize the radio.
During normal operation, sometime for unknown reason, radio encounter
problem and can not recover by itself; the best way to
recover from it is to reset and re-tune the radio. Currently, there is
no RF reset command available, but since radio will get reset when
switching channel, use internal hw scan request to force radio
reset and get back to normal operation state.
The internal hw scan will only perform passive scan on the first
available channel (not the channel being used) in associated state. The
request should be ignored if already performing scan operation or STA is
not in associated state.
Also include an "internal_scan" debugfs file to help trigger the
internal scan from user mode.
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add support to configure missed beacon threshold, by default, if receive
"missed beacon" notification from uCode and has more than 5 consecutive
beacon missed, then perform sensitivity calibration; with this change,
allow user to adjust the missed beacon threshold from debugfs in case
more sensitivity calibration required for better performance in noisy
environment
The default value (=5) should be good enough for the normal condition,
but for very noisy environment, more sensitivity calibration could help
improve the throughput, so by setting the missed beacon threshold to
lower number, user might experience better performance result.
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch introduces three macros to work with uc list from net drivers.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Among other changes, this commit:
commit 06d0f0663e
Author: Sujith <Sujith.Manoharan@atheros.com>
Date: Thu Feb 12 10:06:45 2009 +0530
ath9k: Enable Fractional N mode
changed the hw attach code to fix up initialization values only for
dual band devices, however the commit message did not give a reason as
to why this would be useful or necessary.
According to tests by Jorge Boncompte, this breaks at least some
2GHz-only cards, so the code should be changed back to the
unconditional INI fixup.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Reported-by: Jorge Boncompte <jorge@dti2.net>
Cc: stable@kernel.org
Tested-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
There are a few station addresses that are
char *, instead of the normal u8 *; gcc
gives pointer signedness warnings for some
of those, so use u8 * consistently.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
GC is non-existent in netns, so after you hit GC threshold, no new
dst entries will be created until someone triggers cleanup in init_net.
Make xfrm4_dst_ops and xfrm6_dst_ops per-netns.
This is not done in a generic way, because it woule waste
(AF_MAX - 2) * sizeof(struct dst_ops) bytes per-netns.
Reorder GC threshold initialization so it'd be done before registering
XFRM policies.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Obviously, this register had some other impact that is causing
the regression. Either it is masking some other access or needs
to be reset in some path.
Either, way it is best to just revert the change for 2.6.33
This reverts commit 166a0fd4c7.
Signed-off-by: David S. Miller <davem@davemloft.net>
Bruno Prémont found commit 9793241fe9
(vlan: Precise RX stats accounting) added a regression for non
hw accelerated vlans.
[ 26.390576] BUG: unable to handle kernel NULL pointer dereference at (null)
[ 26.396369] IP: [<df856b89>] vlan_skb_recv+0x89/0x280 [8021q]
vlan_dev_info() was used with original device, instead of
skb->dev. Also spotted by Américo Wang.
Reported-By: Bruno Prémont <bonbons@linux-vserver.org>
Tested-By: Bruno Prémont <bonbons@linux-vserver.org>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
"ip xfrm state|policy count" report SA/SP count from init_net,
not from netns of caller process.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pavel Roskin reported a crash in ieee80211_tx_h_select_key():
http://marc.info/?l=linux-wireless&m=126419655108528&w=2
This is a regression from patch "mac80211: move control.hw_key assignment".
Fix it as suggested by Johannes, adding an else statement to make sure
that tx->key is not accessed when it's null.
Compile-tested only.
Reported-by: Pavel Roskin <proski@gnu.org>
Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Kalle Valo <kalle.valo@iki.fi>
Tested-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Y2_HW_WOL_ON/Y2_HW_WOL_OFF should be set and cleared per chip,
not per port. On dual port cards, Y2_HW_WOL_ON should be
enabled if either sky2 port has WOL enabled.
Found while reviewing code for a WOL regression, though this is
probably not the cause of the regression.
Signed-off-by: Mike McCormack <mikem@ring3k.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on patch from Sridhar Samudrala
The following patch fixes the check for IPv6 GSO packet in ixgbevf
driver to use skb_is_gso_v6(). SKB_GSO_DODGY is also set
when packets are forwarded from a guest.
CC: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Found this problem when testing IPv6 from a KVM guest to a remote
host via e1000e device on the host.
The following patch fixes the check for IPv6 GSO packet in Intel
ethernet drivers to use skb_is_gso_v6(). SKB_GSO_DODGY is also set
when packets are forwarded from a guest.
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
No matter whether connection track is enabled, an end host should send
an ICMPv4 "Fragment Reassembly Timeout" message when defrag timeout.
The reasons are following two points:
1. RFC 792 says:
>>>> >> > > If a host reassembling a fragmented datagram cannot complete the
>>>> >> > > reassembly due to missing fragments within its time limit it
>>>> >> > > discards the datagram, and it may send a time exceeded message.
>>>> >> > >
>>>> >> > > If fragment zero is not available then no time exceeded need be
>>>> >> > > sent at all.
>>>> >> > >
>>>> >> > > Read more: http://www.faqs.org/rfcs/rfc792.html#ixzz0aOXRD7Wp
2. Patrick McHardy also agrees with this opinion. :-)
About the discussion of this opinion, refer to http://patchwork.ozlabs.org/patch/41649
The patch fixed the problem like this:
When enabling connection track, fragments are received at PRE_ROUTING HOOK.
If they are failed to reassemble, ip_expire() will be called.
Before sending an ICMP "Fragment Reassembly Timeout" message,
the patch searches router table to get the destination entry only for host type.
The patch has been tested on both host type and route type.
Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After removing the skb_dma_map/unmap calls the exception handling in
igb_tx_map_adv is not correct. The issue is that the count value was not
being correctly handled so as a result we were not rewinding the ring as
back as we should have been.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simpily pass 'struct neigh_table' with seq_file private pointer,
and save one dereference. Proc entry itself isn't interesting.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After netdev_ops compat code HAVE_* macros aren't needed, in fact
they _will_ result in compile breakage for out of tree drivers.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
