When deploying SFQ/IFB here at work, I found the allot management was
pretty wrong in sfq, even changing allot from short to int...
We should init allot for each new flow, not using a previous value found
in slot.
Before patch, I saw bursts of several packets per flow, apparently
denying the default "quantum 1514" limit I had on my SFQ class.
class sfq 11:1 parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 0b 7p requeues 0
allot 11546
class sfq 11:46 parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 0b 1p requeues 0
allot -23873
class sfq 11:78 parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 0b 5p requeues 0
allot 11393
After patch, better fairness among each flow, allot limit being
respected, allot is positive :
class sfq 11:e parent 11:
(dropped 0, overlimits 0 requeues 86)
backlog 0b 3p requeues 86
allot 596
class sfq 11:94 parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 0b 3p requeues 0
allot 1468
class sfq 11:a4 parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 0b 4p requeues 0
allot 650
class sfq 11:bb parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 0b 3p requeues 0
allot 596
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently return for each active SFQ slot the number of packets in
queue. We can also give number of bytes accounted for these packets.
tc -s class show dev ifb0
Before patch :
class sfq 11:3d9 parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 0b 3p requeues 0
allot 1266
After patch :
class sfq 11:3e4 parent 11:
(dropped 0, overlimits 0 requeues 0)
backlog 4380b 3p requeues 0
allot 1212
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* msm-core: (28 commits)
msm: initial framebuffer support
msm: add handling for clocks tagged as CLK_MINMAX
msm: trout: change name of pmdh_clk to mddi_clk
msm: add CLK_MINMAX to pmdh_clk
msm: trout: add gpio_to_irq
msm: iommu: Use the correct memory allocation flag
msm_serial: Remove redundant unlikely()
msm: iommu: Miscellaneous code cleanup
msm: iommu: Support cache-coherent memory access
msm: iommu: Definitions for extended memory attributes
msm: iommu: Kconfig dependency for the IOMMU API
msm: iommu: Check if device is already attached
msm: iommu: Kconfig item for cacheable page tables
msm: iommu: Don't flush page tables if no devices attached
msm: iommu: Mark functions with the right section names
msm: iommu: Support for the 2nd GFX core's IOMMU
msm: iommu: Revise GFX2D0 IOMMU contexts and M2V mappings
msm: iommu: Revise GFX3D IOMMU contexts and M2V mappings
msm: iommu: Use more consistent naming in platform data
msm: iomap: Addresses and IRQs for 2nd GFX core IOMMU
...
* msm-mmc: (33 commits)
mmc: msm_sdcc: Check for only DATA_END interrupt to end a request
mmc: msm_sdcc: Fix bug in PIO mode when data size is not word aligned
mmc: msm_sdcc: Reset SDCC in case of data transfer errors
mmc: msm_sdcc: Add prog done interrupt support
mmc: msm_sdcc: Fix possible circular locking dependency warning
msm: initial framebuffer support
msm: add handling for clocks tagged as CLK_MINMAX
msm: trout: change name of pmdh_clk to mddi_clk
msm: add CLK_MINMAX to pmdh_clk
msm: trout: add gpio_to_irq
msm: iommu: Use the correct memory allocation flag
msm_serial: Remove redundant unlikely()
msm: iommu: Miscellaneous code cleanup
msm: iommu: Support cache-coherent memory access
msm: iommu: Definitions for extended memory attributes
msm: iommu: Kconfig dependency for the IOMMU API
msm: iommu: Check if device is already attached
msm: iommu: Kconfig item for cacheable page tables
msm: iommu: Don't flush page tables if no devices attached
msm: iommu: Mark functions with the right section names
...
The current code checks for both DATA_END and DATA_BLK_END bits in
MCI_STATUS register and ends a request only if both are set at a time.
The hardware doesn't always set DATA_BLK_END when DATA_END is set.
But DATA_END status itself is sufficient condition from hardware that
data transfer is done and hence, check for only DATA_END interrupt in
software to end a request.
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: David Brown <davidb@codeaurora.org>
The current code for PIO doesn't transfer whole data when data size
is not in multiple of 4 bytes. The last few bytes are not written to
the card resulting in no DATAEND interrupt from SDCC. This patch
allows data transfer for non-aligned data size in PIO mode.
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: David Brown <davidb@codeaurora.org>
SDCC uses an asynchronous FIFOs for data synchronization (one for TX
and one for RX). For any error when DPSM (Data path state machine) is
involved the transfer is terminated with the remaining data stuck inside
FIFOs. Reset the controller in case of data errors to ensure that
any left over data in FIFOs is flushed out and DPSM is in good state.
The following problems are observed without this reset functionality -
1. After the card is removed in an unsafe way (removed when there
is an on going data transfer), the card will not be detected upon
its next insertion. This is because the controller wouldn't respond
to few initialization commands.
2. When an error occurs for a data transfer in non-DMA mode, sometimes
we get spurious PIO interrupt after the request is processed.
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: David Brown <davidb@codeaurora.org>
Enable prog done interrupt for stop command(CMD12) that is sent
after a multi-block write(CMD25). The PROG_DONE bit is set when
the card has finished its programming and is ready for next data.
After every write request the card will be polled for ready status
using CMD13. For a multi-block write(CMD25) before sending CMD13,
stop command (CMD12) will be sent. If we enable prog done interrupt
for CMD12, then CMD13 polling can be avoided. The prog done interrupt
means that the card is done with its programming and is ready for
next request.
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: David Brown <davidb@codeaurora.org>
In the context of request processing thread, data mover lock is
acquired after the host lock. In another context, in the completion
handler of data mover the locks are acquired in the reverse order,
resulting in possible circular lock dependency warning. Hence,
schedule a tasklet to process the dma completion so as to avoid
nested locks.
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: David Brown <davidb@codeaurora.org>
create_workqueue is deprecated. The workqueue usage does not seem to
demand any special treatment, so do not set any flags either.
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Acked-by: Tejun Heo <tj@kernel.org>
Restricting the chainmask to 1 for legacy mode disables useful features
such as MRC, and it reduces the available transmit power.
I can't think of a good reason to do this in legacy mode, so let's just
get rid of that code.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The commit 'ath9k_hw: Disable PAPRD for rates with low Tx power' changed
the code that sets the PAPRD rate masks to use only either the HT20 mask
or the HT40 mask. This is wrong, as the hardware can still use HT20 rates
even when configured for HT40, and the operating channel mode does not
affect PAPRD operation.
The register for the HT40 rate mask is applied as a mask on top of the
other registers to selectively disable PAPRD for specific rates on HT40
packets only.
This patch changes the code back to the old behavior which matches the
intended use of these registers. While with current cards this should not
make any practical difference (according to Atheros, the HT20 and HT40
mask should always be equal), it is more correct that way, and maybe
the HT40 mask will be used for some rare corner cases in the future.
Cc: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When an skb is shared, it needs to be duplicated, along with its data buffer.
If the skb does not have enough headroom, using skb_copy might cause the data
buffer to be copied twice (once by skb_copy and once by pskb_expand_head).
Fix this by using skb_clone initially and letting ieee80211_skb_resize sort
out the rest.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
If the skb is not cloned and we don't need any extra headroom, there
is no point in reallocating the skb head.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The change 'mac80211: Fix BUG in pskb_expand_head when transmitting shared skbs'
added a check for copying the skb if it's shared, however the tx info variable
still points at the cb of the old skb
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
wl12xx_get_platform_data() returns an ERR_PTR on failure and it never
returns a NULL.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Kalle Valo <kvalo@adurom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
A previous conversion from semaphoreto mutexes missed the fact that one
of the semaphores was used in interrupt code. Fixed by changing to
a spinlock.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
ath9k channel table for 2Ghz does not seems to initialize the 'band'
parameter.Though it does not seems to cause any visible issue it looks
odd when we initialize the 'band' parameter for 5Ghz channel table while
not so for 2Ghz.
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When rfkill is enabled, ath9k_hw unnecessarily configured the baseband to
turn off based on GPIO input, however that code was hardcoded to GPIO 0
instead of ah->rfkill_gpio.
Since ath9k uses software rfkill anyway, this code is completely unnecessary
and should be removed in case anything else ever uses GPIO 0.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
To improve aggregation length, there should not be more than two fully formed
A-MPDU frames in the hardware queue. To ensure this, the code checks the tx
queue length before forming new A-MPDUs. This can reduce the throughput (or
maybe even starve out A-MPDU traffic) when too many non-aggregated frames are
in the queue.
Fix this by keeping track of pending A-MPDU frames (even when they're sent out
as single frames), but exclude rate control probing frames to improve
performance.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The Mesh Control header only includes 0, 1 or 2 addresses. If there is
one address, it should be interpreted as Address 4. If there are 2,
they are interpreted as Addresses 5 and 6 (Address 4 being the 4th
address in the 802.11 header).
The address extension used to hold up to 3 addresses instead of the current 2.
I'm not sure which draft version changed this, but it is very unlikely that it
will change again given the state of the approval process of this draft. See
section 7.1.3.6.3 in current draft (8.0).
Also, note that the extra address that I'm removing was not being used, so this
change has no effect on over-the-air frame formats. But I thought I better
remove it before someone does start using it.
Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Export the information which antennas are available for configuration as TX or
RX antennas via nl80211.
Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
As has been pointed out by Daniel Halperin some devices (e.g. Intel IWL5100)
can only TX from a subset of RX antennas, so use separate availability masks
for RX and TX.
Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Let path selection frames for protocols other than HWMP be sent to
userspace via NL80211_CMD_REGISTER_FRAME. Also allow userspace to send
and receive mesh path selection frames.
Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Userspace will now be allowed to toggle between the default path
selection algorithm (HWMP, implemented in the kernel), and a vendor
specific alternative. Also in the same patch, allow userspace to add
information elements to mesh beacons. This is accordance with the
Extensible Path Selection Framework specified in version 7.0 of the
802.11s draft.
Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Mesh parameters can be to setup a mesh or to configure it.
This patch renames the ambiguous name mesh_params to mesh_config
in preparation for mesh_setup.
Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Fix runtime warning with backtrace from hostap by removing
netif_stop_queue() call before register_netdev. Tested to work fine on
hostap_pci Prism 2.5.
(This removes a warning about calling netif_stop_queue before
register_netdev is called. -- JWL)
Signed-off-by: Meelis Roos <mroos@linux.ee>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
All rt2x00 drivers except rt2800pci call ieee80211_tx_status() from
a workqueue, which causes "NOHZ: local_softirq_pending 08" messages.
To fix it, add ieee80211_tx_status_ni() similar to ieee80211_rx_ni()
which can be called from process context, and call it from
rt2x00lib_txdone(). For the rt2800pci special case a driver
flag is introduced.
https://bugzilla.kernel.org/show_bug.cgi?id=24892
Signed-off-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
There were several paths that didn't release their locks.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
info->version only has space for 32 characters but my UTS_RELEASE is
"2.6.37-rc6-next-20101217-05817-ge935fc8-dirty" so it doesn't fit.
This is supposed to be the version of the driver, not the kernel
version. This driver doesn't have a version so lets just leave it
blank.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to active notification of the new MCS7832 version by the manufacturer
(Mr. Milton; thanks!) -- quote: "functionality same as MCS7830",
I'm now submitting this patch (on -rc6), intended for networking.git and -stable.
- add MCS7832 USB PID to be able to support this new device variant, too
- add related descriptions
Signed-off-by: Andreas Mohr <andi@lisas.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Flush the routing cache only of entries that match the
network namespace in which the purge event occurred.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Currently there are some info message that is set as error, and an error
message that is set as debug. This patch just fixes it.
Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6:
[media] gspca - sonixj: Better handling of the bridge registers 0x01 and 0x17
[media] gspca - sonixj: Add the bit definitions of the bridge reg 0x01 and 0x17
[media] gspca - sonixj: Set the flag for some devices
[media] gspca - sonixj: Add a flag in the driver_info table
[media] gspca - sonixj: Fix a bad probe exchange
[media] gspca - sonixj: Move bridge init to sd start
[media] bttv: remove unneeded locking comments
[media] bttv: fix mutex use before init (BZ#24602)
[media] Don't export format_by_forcc on two different drivers
Currently, destroy_workqueue() makes the workqueue deny all new
queueing by setting WQ_DYING and flushes the workqueue once before
proceeding with destruction; however, there are cases where work items
queue more related work items. Currently, such users need to
explicitly flush the workqueue multiple times depending on the
possible depth of such chained queueing.
This patch updates the queueing path such that a work item can queue
further work items on the same workqueue even when WQ_DYING is set.
The flush on destruction is automatically retried until the workqueue
is empty. This guarantees that the workqueue is empty on destruction
while allowing chained queueing.
The flush retry logic whines if it takes too many retries to drain the
workqueue.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
hna_local_fill_buffer must return the number of added hna entries and
not the last checked hash bucket.
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove "pktgen: " prefix string from one pr_info.
pr_fmt adds it, so this is a duplicate.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename log_level to bfa_log_level to make the global variable more bfa
specific and avoid clashes with other drivers which was causing a
build failure.
Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Pawel reported a panic related to handling shared skbs in ixgbe
incorrectly. So we need to revert my previous patch to work around
this bug. Instead of reverting the patch completely, I just revert
the essential lines, so we can add the previous optimization
back more easily in future.
commit 3511c9132f
Author: Changli Gao <xiaosuo@gmail.com>
Date: Sat Oct 16 13:04:08 2010 +0000
net_sched: remove the unused parameter of qdisc_create_dflt()
Reported-by: Pawel Staszewski <pstaszewski@itcare.pl>
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'fbdev-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/lethal/fbdev-2.6:
OMAP: OMAPFB: disable old omapfb for OMAP4 builds
OMAP: DSS: VRAM: Align start & size of vram to 2M
* 's5p-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
ARM: S5PV210: update MAX8998 platform data to get rid of WARN()
ARM S3C24XX: Fix compilation of PM code for S3C2416
ARM: S3C24XX: Fix CONFIG_S3C_DEV_NAND Kconfig entry
* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
cciss: fix cciss_revalidate panic
block: max hardware sectors limit wrapper
block: Deprecate QUEUE_FLAG_CLUSTER and use queue_limits instead
blk-throttle: Correct the placement of smp_rmb()
blk-throttle: Trim/adjust slice_end once a bio has been dispatched
block: check for proper length of iov entries earlier in blk_rq_map_user_iov()
drbd: fix for spin_lock_irqsave in endio callback
drbd: don't recvmsg with zero length
The cnt32_to_63 algorithm relies on proper counter data evaluation
ordering to work properly. This was missing from the provided
documentation.
Let's augment the documentation with the missing usage constraint and
fix the only instance that got it wrong.
Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
