sorry, just realized skb_dequeue() will call spin_lock_irqsave() inside.
please ignore this question.
On Sat, Sep 9, 2017 at 9:28 AM, dztemplor . wrote:
> Hi, everyone.
> I'm a newbie in mac80211 and have a question about synchronization for
> sdata->skb_queue.
>
Hi, everyone.
I'm a newbie in mac80211 and have a question about synchronization for
sdata->skb_queue.
sdata->skb_queue is shared between ieee80211_iface_work() and rx
handlers like ieee80211_rx_h_action()
ieee80211_iface_work() runs in process context.
ieee80211_rx_h_action() runs in softirq
On Mon, Sep 4, 2017 at 6:19 AM, Johannes Berg wrote:
> On Fri, 2017-09-01 at 13:07 -0700, Thomas Pedersen wrote:
>> On Thu, Aug 31, 2017 at 11:30 PM, Greg Maitz
>> wrote:
>> > Hi guys,
>> >
>> > I'm seeing a problem when I work on the wireless mesh
Hi,
Could have used a 'v2' in the subject, but hopefully that doesn't bother
Kalle too much.
On Fri, Sep 08, 2017 at 02:02:43AM +0530, Ganapathi Bhat wrote:
> If driver is loaded with 'mfg_mode' enabled, then the sending
> commands are not allowed. So, skip sending commands, to firmware
> in
In brcmf_p2p_notify_rx_mgmt_p2p_probereq(), chanspec is assigned before
the length of rxframe is validated. This could lead to uninitialized
data being printed in a debug message. Since we already have a
perfectly good endian-swapped copy of rxframe->chanspec in ch.chspec,
and ch.chspec is not
The length of the data in the received skb is currently passed into
brcmf_fweh_process_event() as packet_len, but this value is not checked.
event_packet should be followed by DATALEN bytes of additional event
data. Ensure that the received packet actually contains at least
DATALEN bytes of
These were suggested by the Chrome OS security team[1]. Compile-tested
only.
[1] http://crosreview.com/656260
Kevin Cernekee (3):
brcmfmac: Avoid possible out-of-bounds read
brcmfmac: Don't print out-of-bounds event data
brcmfmac: Add check for short event packets
From: Kalle Valo
Date: Fri, 08 Sep 2017 18:50:01 +0300
> few fixes to net tree for 4.14. Note that this pull request contains the
> iwlwifi fix Linus hopes to have by end of the merge window. Please let
> me know if there are any problems.
Pulled, thanks for following up
From: Iok Chan
Sent: Saturday, 9 September 2017 2:11 a.m.
Subject: OUTLOOK SLOW PERFORMANCE
Opening Statement:
Our Outlook platform is currently operational, but running very slower than
usual. Support is currently investigating the issue.Please you need
Use IEEE80211_RATE_MANDATORY_G instead of IEEE80211_RATE_MANDATORY_B to get
all mandatory rates in 2.4 GHz band. It is safe to do so because ERP
mandatory rates are a superset of HR/DSSS mandatory rates. Also limit to
OFDM rates for 10 MHz and 5 MHz channels as originally intended by commit
Hi Dave,
few fixes to net tree for 4.14. Note that this pull request contains the
iwlwifi fix Linus hopes to have by end of the merge window. Please let
me know if there are any problems.
Kalle
The following changes since commit 8e0deed92406d93ae0365cb8a6134db5721e7aca:
tipc: remove
From: Colin Ian King
The assignment of dev is dereferencing adapter before adapter has
been null checked, potentially leading to a null pointer dereference.
Fix this by simply moving the assignment of dev to a later point
after the sanity null check of adapter.
On Fri, 2017-09-08 at 17:07 +0200, Benjamin Beichler wrote:
>
> Am 8. September 2017 16:19:20 MESZ schrieb Johannes Berg psolutions.net>:
> > On Fri, 2017-09-08 at 16:11 +0200, Benjamin Beichler wrote:
> > > The ops field is zero initialized, therefore parallel ops is
> > > already
Am 8. September 2017 16:19:20 MESZ schrieb Johannes Berg
:
>On Fri, 2017-09-08 at 16:11 +0200, Benjamin Beichler wrote:
>> The ops field is zero initialized, therefore parallel ops is already
>> false.
>
>Therefore this patch is completely pointless?
Sorry my first
Uh. Is there anything you can do not to mangle your replies? I don't
even know what you're replying to etc.
> AFAIK, you can change the MAC-Address of an interface on
> the phy, but not the perm address of the phy. But for frames sent to
> a wmediumd always the perm address is used to identify
>> Since the radio index is already a valid NL attribute, this patch >> simply
>> try to read it and create the radio specific to this index >>
by the previously used scheme. >> >> Since this allows to create radios
out of row, we need to search >> for a free index for a new radio, when
the
On Fri, 2017-09-08 at 16:11 +0200, Benjamin Beichler wrote:
> Since the radio index is already a valid NL attribute, this patch
> simply
> try to read it and create the radio specific to this index by the
> previously used scheme.
>
> Since this allows to create radios out of row, we need to
This patch series includes our efforts for more sophisticated simulations for
wifi-networks.
Especially Patch 2 and 4 add missing functionality.
The patch 1 adds an obvious performance improvement for many radios, since for
every received frame a linear search through all radios is done.
On Fri, 2017-09-08 at 16:11 +0200, Benjamin Beichler wrote:
> For correct interpretation of a tx rate, the corresponding rate flags
> are needed (e.g. whether a HT-MCS rate or a legacy rate) and
> moreover for more correct simulation the other infos of the flags are
> important (like short-GI).
On Fri, 2017-09-08 at 16:11 +0200, Benjamin Beichler wrote:
> Use hashlist to improve lookup speed for every received NL-message,
> especially for higher counts of radios, like 100 or more. The
> stringhash
> should be neglible for 6 Bytes MAC-Address, could be improved by
> using
> cast to 64bit
On Fri, 2017-09-08 at 16:11 +0200, Benjamin Beichler wrote:
> The ops field is zero initialized, therefore parallel ops is already
> false.
Therefore this patch is completely pointless?
johannes
For correct interpretation of a tx rate, the corresponding rate flags are
needed (e.g. whether a HT-MCS rate or a legacy rate) and
moreover for more correct simulation the other infos of the flags are important
(like short-GI). Keeping compability, the flags
are not integrated into the existing
Use hashlist to improve lookup speed for every received NL-message,
especially for higher counts of radios, like 100 or more. The stringhash
should be neglible for 6 Bytes MAC-Address, could be improved by using
cast to 64bit integer and hash this instead.
For a more reliable implementation of
Since the radio index is already a valid NL attribute, this patch simply
try to read it and create the radio specific to this index by the
previously used scheme.
Since this allows to create radios out of row, we need to search for a
free index for a new radio, when the index is not present as
The ops field is zero initialized, therefore parallel ops is already false.
This implicates that the netlink callbacks are not
processed in parallel. Maybe this could be utilized to reduce locking overhead
or maybe also parallel ops could be implemented.
Signed-off-by: Benjamin Beichler
Hello,
My name is Cristian Aravena Romero, [1]Free Software enthusiast in
Chile. My _Notebook_ has a WiFi card [2]"Intel® Centrino® Advanced-N
6235" running from version 3.2 of the Linux kernel and above; and also
occupies the firmware "iwlwifi-6000g2b-ucode-18.168.6.1.tgz". The
complete
I assume that a blank efuse comes with all ones, thus I did not bother
recognizing other possible junk values. This matches 100% of dongles
I've seen (a single Gembird 8192eu).
Signed-off-by: Adam Borowski
---
v2: strncmp("goofy string") -> memchr_inv()
Stefano Brivio
On Fri, 2017-09-08 at 12:10 +0200, Richard Schütz wrote:
>
> This would leave us with 1, 2, 5.5 and 11 Mb/s for ERP PHYs again
> when it really should be 1, 2, 5.5, 6, 11, 12 and 24 Mb/s.
>
Yes, you're right - I had thought about that but forgot. The places
that check it would have to be
Am 08.09.2017 um 11:03 schrieb Johannes Berg:
On Fri, 2017-09-08 at 10:43 +0200, Richard Schütz wrote:
Am 08.09.2017 um 08:55 schrieb Johannes Berg:
On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
Use IEEE80211_RATE_MANDATORY_G instead of
IEEE80211_RATE_MANDATORY_B
for comparison to
From: Johannes Berg
This was created using the following spatch:
@find@
type S;
expression M, M2;
position p;
@@
offsetof(S, M) + sizeof(M2)@p
@script:python@
m << find.M;
m2 << find.M2;
@@
if not m2.endswith('-> ' + m):
Hi,
On Friday, September 8, 2017 10:53:37 AM CEST Richard Schütz wrote:
> Am 08.09.2017 um 10:43 schrieb Richard Schütz:
> > Am 08.09.2017 um 08:55 schrieb Johannes Berg:
> >> On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
> >>> Use IEEE80211_RATE_MANDATORY_G instead of
Ian W MORRISON wrote:
> The firmware feature check introduced for multi-scheduled scan is also
> failing for bcm4345 devices resulting in a firmware crash.
> The reason for this crash has not yet been root cause so this patch avoids
> the feature check for those device as
Hi Ian,
please don't send me email privately, instead always send the questions
to the list. You get better answers faster and I get less email :)
Ian W MORRISON writes:
> With the below patch how can I track when it has been merged into 4.13
> so I can check and start
From: Luca Coelho
We always must set the status to what we consider success before
calling iwl_mvm_send_cmd_status() (also iwl_mvm_send_cmd_pdu_status()
which calls it). Fix a few places where initialization is missing.
Signed-off-by: Luca Coelho
From: Naftali Goldstein
Add a new state to enum iwl_mvm_agg_state, which is used between
queueing the work that starts tx aggregations and actually starting that
work (changing to state IWL_AGG_STARTING).
This solves a race where ieee80211_start_tx_ba_session is
From: Sara Sharon
The condition to check if reorder buffer ran out of
space is faulty, as it takes into account only the
NSSN.
In case the head SN was too far behind the reorder
buffer should move forward, regardless of the NSSN
status.
This caused the driver to release
From: Luca Coelho
We always need to initialize the status argument to the success case
before calling iwl_mvm_send_cmd_status() or
iwl_mvm_send_cmd_pdu_status() (which calls the former) otherwise we
may get an uninitialized value back. In this case, we use
From: Luca Coelho
We were ignoring the FIF_ALLMULTI flag when setting the multicast
addresses with MCAST_FILTER_CMD. Check if this flag is set and enable
pass_all accordingly. We also need to set the count to 0 if pass_all
is enable so we don't pass addresses to the
From: Luca Coelho
The MCAST_FILTER_CMD can get quite large when we have many mcast
addresses to set (we support up to 255). So the command should be
send as NOCOPY to prevent a warning caused by too-long commands:
WARNING: CPU: 0 PID: 9700 at
From: Avraham Stern
AP interfaces now send all non-bufferable frames using the broadcast
station. Thus allow them to use the probe queue and don't warn about
it.
Fixes: eb045e6e0389 ("iwlwifi: mvm: Avoid deferring non bufferable frames")
Signed-off-by: Avraham Stern
From: David Spinadel
When starting wowlan mac80211 requests flush w/o vif
and we ignore this request. As a result some packets
stay stuck in the queue and it may end up with a queue
hang.
Allow the driver to flush queues even if station isn't
specified.
Signed-off-by:
From: Avraham Stern
iwl_mvm_start_mac_queues() takes a bitmap of the queues to wake.
When deferred tx is purged, set the bit of the hw_queue so
the correct queue will be waken up.
Fixes: 7e39a00d5931 ("iwlwifi: mvm: start mac queues when deferred tx frames
are purged")
From: Matt Chen
If we set disconnect on wowlan and run suspend/resume, will run
into:
...snipped
iwlwifi :01:00.0: Failed to load firmware chunk!
iwlwifi :01:00.0: Could not load the [0] uCode section
iwlwifi :01:00.0: Failed to start INIT ucode: -110
iwlwifi
From: Luca Coelho
Hi,
Here is my first batch of fixes intended for 4.14. Since it is very
early in the release process, I was a bit liberal with the fixes I'm
sending, but they all fix real issues.
These are the fixes:
* A couple of bugzilla bugs related to
On Fri, 2017-09-08 at 10:43 +0200, Richard Schütz wrote:
> Am 08.09.2017 um 08:55 schrieb Johannes Berg:
> > On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
> > > Use IEEE80211_RATE_MANDATORY_G instead of
> > > IEEE80211_RATE_MANDATORY_B
> > > for comparison to get all mandatory rates in
On Fri, 2017-09-08 at 10:43 +0200, Richard Schütz wrote:
> > Hmm, I guess technically you're correct, since 11b added what's now
> > Clause 16 (was Clause 18 at the time), and that has all of them
> > mandatory? But perhaps this was actually intended for Clause 15
> > compatibility?
>
>
Am 08.09.2017 um 10:43 schrieb Richard Schütz:
Am 08.09.2017 um 08:55 schrieb Johannes Berg:
On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
Use IEEE80211_RATE_MANDATORY_G instead of IEEE80211_RATE_MANDATORY_B
for comparison to get all mandatory rates in 2.4 GHz band. It is safe
to do
Am 08.09.2017 um 08:54 schrieb Johannes Berg:
On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
According to IEEE Std 802.11-2016 (16.2.3.4 Long PHY SIGNAL field)
all of
the following rates are mandatory for a HR/DSSS PHY: 1 Mb/s, 2 Mb/s,
5.5 Mb/s and 11 Mb/s. Set
Am 08.09.2017 um 08:55 schrieb Johannes Berg:
On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
Use IEEE80211_RATE_MANDATORY_G instead of IEEE80211_RATE_MANDATORY_B
for comparison to get all mandatory rates in 2.4 GHz band. It is safe
to do so because ERP mandatory rates are a superset
On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
> Use IEEE80211_RATE_MANDATORY_G instead of IEEE80211_RATE_MANDATORY_B
> for comparison to get all mandatory rates in 2.4 GHz band. It is safe
> to do so because ERP mandatory rates are a superset of HR/DSSS
> mandatory rates.
This I don't
On Thu, 2017-09-07 at 17:47 +0200, Richard Schütz wrote:
> According to IEEE Std 802.11-2016 (16.2.3.4 Long PHY SIGNAL field)
> all of
> the following rates are mandatory for a HR/DSSS PHY: 1 Mb/s, 2 Mb/s,
> 5.5 Mb/s and 11 Mb/s. Set IEEE80211_RATE_MANDATORY_B flag for all of
> these instead of
51 matches
Mail list logo