Jiri Benc [Sat, 28 Mar 2015 18:13:22 +0000 (19:13 +0100)]
ipvlan: fix addr hash list corruption
When ipvlan interface with IP addresses attached is brought down and then
deleted, the assigned addresses are deleted twice from the address hash
list, first on the interface down and second on the link deletion.
Similarly, when an address is added while the interface is down, it is added
second time once the interface is brought up.
When the interface is down, the addresses should be kept off the hash list
for performance reasons. Ensure this is true, which also fixes the double add
problem. To fix the double free, check whether the address is hashed before
removing it.
Alexander Duyck [Fri, 27 Mar 2015 21:14:22 +0000 (14:14 -0700)]
fib_trie: Cleanup ip_fib_net_exit code path
While fixing a recent issue I noticed that we are doing some unnecessary
work inside the loop for ip_fib_net_exit. As such I am pulling out the
initialization to NULL for the locally stored fib_local, fib_main, and
fib_default.
In addition I am restoring the original code for flushing the table as
there is no need to split up the fib_table_flush and hlist_del work since
the code for packing the tnodes with multiple key vectors was dropped.
The issue was that as a part of exiting the default rules were being
deleted which resulted in the local trie being unmerged. By moving the
freeing of the FIB tables up we can avoid the unmerge since there is no
local table left when we call the fib4_rules_exit function.
Fixes: 0ddcf43d5d4a ("ipv4: FIB Local/MAIN table collapse") Reported-by: Cong Wang <[email protected]> Signed-off-by: Alexander Duyck <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Here's another set of Bluetooth & 802.15.4 patches for 4.1:
- New API to control LE advertising data (i.e. peripheral role)
- mac802154 & at86rf230 cleanups
- Support for toggling quirks from debugfs (useful for testing)
- Memory leak fix for LE scanning
- Extra version info reading support for Broadcom controllers
Please let me know if there are any issues pulling. Thanks.
====================
David S. Miller [Tue, 31 Mar 2015 16:01:21 +0000 (12:01 -0400)]
Merge branch 'ptp-2038'
Fixed two warnings in e1000e and igb, when switching to timespec64
some printf formats started to not match. In theses cases actually
the new type is __kernel_time_t which is __kernel_long_t which
unfortunately can be either "long" or "long long". So to solve
this I cases the arguments to "long long". -DaveM
Richard Cochran says:
====================
ptp: get ready for 2038
This series converts the core driver methods of the PTP Hardware Clock
(PHC) subsystem to use the 64 bit version of the timespec structure,
making the core API ready for the year 2038.
In addition, I reviewed how each driver and device represents the time
value at the hardware register level. Most of the drivers are ready,
but a few will need some work before the year 2038, as shown:
The commit log messages document how each driver is ready or why it is
not ready. For patch 15, I could not easily find out the hardware
representation of the time value, and so the SFC maintainers will have
to review their low level code in order to resolve any remaining
issues.
* ChangeLog
** V3
- dp83640: use timespec64 throughout per Arnd's suggestion
- tilegx: use timespec64 throughout per Chris' suggestion
- add Jeff's acked-bys
** V2
- use the new methods in the posix clock code right away (patch #3)
====================
Richard Cochran [Sun, 29 Mar 2015 21:12:13 +0000 (23:12 +0200)]
ptp: remove 32 bit get/set methods.
All of the PHC drivers have been converted to the new methods. This patch
converts the three remaining callers within the core code and removes the
older methods for good. As a result, the core PHC code is ready for the
year 2038. However, some of the PHC drivers are not quite ready yet.
Richard Cochran [Sun, 29 Mar 2015 21:12:10 +0000 (23:12 +0200)]
ptp: dp83640: convert to the 64 bit get/set time methods.
This device stores the number of seconds in a 32 bit register, and the
stored value is unsigned. Therefore this driver and device are ready
for the year 2038. However, more work will be needed prior to 2106.
Richard Cochran [Sun, 29 Mar 2015 21:12:09 +0000 (23:12 +0200)]
ptp: tilegx: convert to the 64 bit get/set time methods.
This driver is 64 bit only, and so this driver and device are ready
for 2038. This patch changes the driver to the new PHC and also
carries the timespec64 parameter on out to the gxio_mpipe_get-
set_timestamp functions, making explicit the fact that the tv_sec
field is 64 bits wide.
Richard Cochran [Sun, 29 Mar 2015 21:12:03 +0000 (23:12 +0200)]
ptp: igb: convert to the 64 bit get/set time methods.
For the 82576, the driver's clock is implemented using a timecounter,
and so with this patch that device is ready for the year 2038.
However, in the case of the i210, the device stores the number of
seconds in a 32 bit register. Therefore, more work is needed on this
driver before the year 2038 comes around.
Richard Cochran [Sun, 29 Mar 2015 21:11:51 +0000 (23:11 +0200)]
ptp: introduce get/set time methods with explicit 64 bit seconds.
Converting the PHC drivers over to the new methods is one step along the
way to making them ready for 2038. Once all the drivers are up to date,
then the old methods will be removed.
Mike Snitzer [Mon, 30 Mar 2015 17:39:09 +0000 (13:39 -0400)]
block: fix blk_stack_limits() regression due to lcm() change
Linux 3.19 commit 69c953c ("lib/lcm.c: lcm(n,0)=lcm(0,n) is 0, not n")
caused blk_stack_limits() to not properly stack queue_limits for stacked
devices (e.g. DM).
Fix this regression by establishing lcm_not_zero() and switching
blk_stack_limits() over to using it.
DM uses blk_set_stacking_limits() to establish the initial top-level
queue_limits that are then built up based on underlying devices' limits
using blk_stack_limits(). In the case of optimal_io_size (io_opt)
blk_set_stacking_limits() establishes a default value of 0. With commit 69c953c, lcm(0, n) is no longer n, which compromises proper stacking of
the underlying devices' io_opt.
Johan Hedberg [Mon, 30 Mar 2015 20:21:02 +0000 (23:21 +0300)]
Bluetooth: Refactor HCI request variables into own struct
In order to shrink the size of bt_skb_cb, this patch moves the HCI
request related variables into their own req_ctrl struct. Additionall
the L2CAP and HCI request structs are placed inside the same union since
they will never be used at the same time for the same skb.
Johan Hedberg [Mon, 30 Mar 2015 20:21:01 +0000 (23:21 +0300)]
Bluetooth: Refactor L2CAP variables into l2cap_ctrl
We're getting very close to the maximum possible size of bt_skb_cb. To
prepare to shrink the struct with the help of a union this patch moves
all L2CAP related variables into the l2cap_ctrl struct. To later add
other 'ctrl' structs the L2CAP one is renamed simple 'l2cap' instead
of 'control'.
Linus Torvalds [Mon, 30 Mar 2015 18:08:37 +0000 (11:08 -0700)]
Merge branch 'for-4.0-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata
Pull libata fixes from Tejun Heo:
"Nothing exciting. Two patches to update queued trim blacklist"
* 'for-4.0-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata:
libata: Blacklist queued TRIM on Samsung SSD 850 Pro
libata: Update Crucial/Micron blacklist
Peter Ujfalusi [Fri, 27 Mar 2015 11:35:52 +0000 (13:35 +0200)]
dmaengine: omap-dma: Fix memory leak when terminating running transfer
In omap_dma_start_desc the vdesc->node is removed from the virt-dma
framework managed lists (to be precise from the desc_issued list).
If a terminate_all comes before the transfer finishes the omap_desc will
not be freed up because it is not in any of the lists and we stopped the
DMA channel so the transfer will not going to complete.
There is no special sequence for leaking memory when using cyclic (audio)
transfer: with every start and stop of a cyclic transfer the driver leaks
struct omap_desc worth of memory.
Free up the allocated memory directly in omap_dma_terminate_all() since the
framework will not going to do that for us.
Petr Kulhavy [Fri, 27 Mar 2015 11:35:51 +0000 (13:35 +0200)]
dmaengine: edma: fix memory leak when terminating running transfers
If edma_terminate_all() was called while a transfer was running (i.e. after
edma_execute() but before edma_callback()) the echan->edesc was not freed.
This was due to the fact that a running transfer is on none of the
vchan lists: desc_submitted, desc_issued, desc_completed (edma_execute()
removes it from the desc_issued list), so the vchan_dma_desc_free_list()
called at the end of edma_terminate_all() didn't find it and didn't free it.
This bug was found on an AM1808 based hardware (very similar to da850evm,
however using the second MMC/SD controller), where intense operations on the SD
card wasted the device 128MB RAM within a couple of days.
Peter Ujfalusi:
The issue is even more severe since it affects cyclic (audio) transfers as
well. In this case starting/stopping audio will results memory leak.
Linus Torvalds [Mon, 30 Mar 2015 16:14:41 +0000 (09:14 -0700)]
Merge tag 'gpio-v4.0-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio
Pull late GPIO fixes from Linus Walleij:
"Here are the (hopefully) last GPIO fixes for v4.0. Nothing
controversial whatsoever, just fixes:
- syscon GPIO fix for Keystone DSP GPIOs
- pin number translation fix for ACPI GPIO
- a smallish compiler warning fix on the mpc8xxx driver"
* tag 'gpio-v4.0-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio:
gpio: syscon: reduce message level when direction reg offset not in dt
gpiolib: translate pin number in GPIO ACPI callbacks
gpio: mpc8xxx: remove __initdata annotation for mpc8xxx_gpio_ids[]
drm/i915: Use batch length instead of object size in command parser
Testcase: igt/gem_exec_nop #ivb,byt,hsw Signed-off-by: Chris Wilson <[email protected]> Signed-off-by: Daniel Vetter <[email protected]>
[cherry picked from commit 743e78c1d726d875b98ff9689cc77c4d3d5d9ae2
from drm-intel-next because 4.0 seems to be affected by this too,
despite that the obvious culprit is definitely not in 4.0. Whatever,
if fixes a bug. Signed-off-by: Daniel Vetter <[email protected]>
Johannes Berg [Mon, 30 Mar 2015 13:09:20 +0000 (15:09 +0200)]
mac80211: set QoS capability before changing station state
In the upcoming fast-xmit patch, changing station state will
build a header cache based on the station's capabilities, and
as the QoS capability (sta.wme) impacts the header, it needs
to be set before.
In case of wide bandwidth (wider than 20MHz) used by IBSS,
scan all channels in chandef to be able to find neighboring
IBSS netwqworks that use the same overall channels but a different
control channel.
Johannes Berg [Sat, 21 Mar 2015 08:13:45 +0000 (09:13 +0100)]
mac80211: factor out station lookup from ieee80211_build_hdr()
In order to look up the RA station earlier to implement a TX
fastpath, factor out the lookup from ieee80211_build_hdr().
To always have a valid station pointer, also move some of the
checks into the new function.
Johannes Berg [Sat, 21 Mar 2015 07:09:55 +0000 (08:09 +0100)]
mac80211: make sta.wme indicate whether QoS is used
Indicating just the peer's capability is fairly pointless
if the local device doesn't support it. Make the variable
track both combined, and remove the 'local support' check
in the TX path.
Johannes Berg [Sat, 21 Mar 2015 06:41:04 +0000 (07:41 +0100)]
mac80211: send AP probe as unicast again
Louis reported that a static checker was complaining that
the 'dst' variable was set (multiple times) but not used.
This is due to a previous commit having removed the usage
(apparently erroneously), so add it back.
Fixes: a344d6778a98 ("mac80211: allow drivers to support NL80211_SCAN_FLAG_RANDOM_ADDR") Reported-by: Louis Langholtz <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
Johannes Berg [Mon, 23 Mar 2015 14:08:14 +0000 (17:08 +0300)]
mac80111: aes_gcm: clean up ieee80211_aes_gcm_key_setup_encrypt()
This code is written using an anti-pattern called "success handling"
which makes it hard to read, especially if you are used to normal kernel
style. It should instead be written as a list of directives in a row
with branches for error handling.
(Basically copied from Dan's previous patch for CCM)
This code is written using an anti-pattern called "success handling"
which makes it hard to read, especially if you are used to normal kernel
style. It should instead be written as a list of directives in a row
with branches for error handling.
Jouni Malinen [Mon, 23 Mar 2015 13:41:15 +0000 (15:41 +0200)]
mac80211: Fix misplaced return in AES-GMAC key setup
Commit 8ade538bf39b ("mac80111: Add BIP-GMAC-128 and BIP-GMAC-256
ciphers") had the success return in incorrect place before the
crypto_aead_setauthsize() call which practically ended up skipping that
call unconditionally.
The missing call did not actually change any functionality since
GMAC_MIC_LEN (16) is identical to the maxauthsize in gcm(aes) and as
such, the default value used for the authsize parameter.
Tom Gundersen [Wed, 18 Mar 2015 10:13:39 +0000 (11:13 +0100)]
cfg80211: pass name_assign_type to rdev_add_virtual_intf()
This will expose in /sys whether the ifname of a device is set by
userspace or generated by the kernel. The latter kind (wlanX, etc)
is not deterministic, so userspace needs to rename these devices
to names that are guaranteed to stay the same between reboots. The
former, however should never be renamed, so userspace needs to be
able to reliably tell the difference.
Similar functionality was introduced for the rtnetlink core in
commit 5517750f058e ("net: rtnetlink - make create_link take name_assign_type")
Zhaoyang Liu [Mon, 23 Mar 2015 14:20:56 +0000 (19:50 +0530)]
mwifiex: recover from skb allocation failures during RX
This patch adds recovery mechanism for SDIO RX during SKB allocation
failures.
For allocation failures during multiport aggregation, we skip and drop RX
packets.
For single port read case, we will use preallocated card->mpa_rx.buf to
complete cmd53 read.
Now we terminate SDIO operations only upon cmd53 failures.
Dan Carpenter [Mon, 23 Mar 2015 09:35:51 +0000 (12:35 +0300)]
mwifiex: remove an unneede NULL check in mwifiex_init_adapter()
"adapter->sleep_cfm" is always non-NULL at this point. Static checkers
complain that we already dereference it at the start of the function
when we do:
Johannes Berg [Mon, 9 Mar 2015 10:13:04 +0000 (11:13 +0100)]
mac80211: reject aggregation sessions with non-HT peers
If a peer or some local agent (rate control, ...) decides to start
an aggregation session but doesn't support HT (which also implies
QoS), reject it.
This is mostly a corner case as such peers normally won't try to
use block-ack sessions and rate control wouldn't start them, but
technically QoS stations could request it according to the spec.
However, since drivers don't really support such non-HT sessions
it's better to reject them.
Also, while at it, move the tracing for TX sessions earlier so it
captures the error cases as well.
Maithili Hinge [Mon, 23 Mar 2015 06:01:25 +0000 (23:01 -0700)]
mwifiex: Fix issue in the SDIO reset path of mwifiex.
SDIO reset was not happening properly on mwifiex as cancel_work_sync
in mwifiex_sdio_remove used to kill the calling work function itself.
Due to this, the interface was not getting removed and card was not
getting added again. Reset work function has been made independent
of adapter variable and cancel_work_sync has been moved to cleanup
function.
Oleksij Rempel [Sun, 22 Mar 2015 18:29:51 +0000 (19:29 +0100)]
ath9k: add new function ath9k_hw_read_array
REG_READ generate most overhead on usb bus. It send and read micro packages
and reduce usb bandwidth. To reduce this overhead we should read in batches.
Oleksij Rempel [Sun, 22 Mar 2015 18:29:46 +0000 (19:29 +0100)]
ath9k_htc: add new WMI_REG_RMW_CMDID command
Since usb bus add extra delay on each request, a command
with read + write requests is too expensive. We can dramtically
reduce usb load by moving this command to firmware.
In my tests, this patch will reduce channel scan time
for about 5-10 seconds.
rt2800usb: check Autorun mode on FW load only once
Seems H/W report correctly firmware Autorun value only at initialization
stage. When we close interface and open it again, Autorun value is 0 and
we try to load firmware what kills the device. To fix clear
REQUIRE_FIRMWARE firmware flag, to do not load firmware again, once we
discover Autorun mode.
Arik Nemtsov [Wed, 18 Mar 2015 06:46:08 +0000 (08:46 +0200)]
cfg/mac80211: add regulatory classes IE during TDLS setup
Seems Broadcom TDLS peers (Nexus 5, Xperia Z3) refuse to allow TDLS
connection when channel-switching is supported but the regulatory
classes IE is missing from the setup request.
Add a chandef to reg-class translation function to cfg80211 and use it
to add the required IE during setup. For now add only the current
regulatory class as supported - it is enough to resolve the
compatibility issue.
David Spinadel [Tue, 17 Mar 2015 17:58:38 +0000 (19:58 +0200)]
mac80211: stop scan before connection
Stop scan before authentication or association to make sure
that nothing interferes with connection flow.
Currently mac80211 defers RX auth and assoc packets (among other ones)
until after the scan is complete, so auth during scan is likely to fail
if scan took too much time.
mac80211: convert rssi_callback() to event_callback()
We will be able to add more events, such as MLME events and
others. The low level driver may be interested in knowing
about these events to dump firmware data upon failures, or
to change parameters in case connection attempts fail etc...
Johannes Berg [Thu, 12 Mar 2015 18:28:31 +0000 (19:28 +0100)]
mac80211: agg-tx: avoid sending DelBA with sta->lock held
The rate control locking caused a potential deadlock here due to the
locks being acquired in different orders, so that change cannot yet
be applied. However, there's no fundamental reason for this code to
hold the sta->lock while transmitting frames.
Clearly it's better not to hold the lock for longer periods of time,
which can happen here since we call all the way down to the driver.
Change the code a bit to not hold it while doing that.
Alexander Bondar [Thu, 26 Mar 2015 09:07:35 +0000 (11:07 +0200)]
iwlwifi: mvm: Clean up UMAC scan UIDs in the reset and drv_stop flows
In the reset flow, the driver cancels ongoing scan and sends scan
complete notification to mac80211. However it does not clean its UID.
Add cleaning scan UID for the ongoing scan. Loop over all other UIDs
to make sure there's nothing left there and warn if any is found.
Eran Harary [Tue, 24 Mar 2015 08:59:46 +0000 (10:59 +0200)]
iwlwifi: 8000: change PNVM in case it doesn't match to the HW step
There is a strong relationship between the NVM version and
the hardware step. Enforce that in the driver in case the
default NVM on the platform is the wrong one.
Dave Airlie [Mon, 30 Mar 2015 02:17:27 +0000 (12:17 +1000)]
Merge branch 'drm-fixes-4.0' of git://people.freedesktop.org/~agd5f/linux into drm-fixes
A few more fixes for 4.0 for radeon. Sorry for the delay, I was
a little under the weather this week and time got away from me.
* 'drm-fixes-4.0' of git://people.freedesktop.org/~agd5f/linux:
drm/radeon: programm the VCE fw BAR as well
drm/radeon: always dump the ring content if it's available
radeon: Do not directly dereference pointers to BIOS area.
drm/radeon/dpm: fix 120hz handling harder
projects / linux.git / log