Luuk Paulussen [Tue, 22 Dec 2015 20:51:14 +0000 (09:51 +1300)]
staging: octeon: Fix logic for waking octeon ethernet tx queue.
Only wake tx queue when driver queue is back within bounds.
The logic here was just reenabling the queue when any buffers had been
freed. the queue was stopped whenever the length exceeded 1000
(MAX_OUT_QUEUE_DEPTH), but then was essentially immediately started again.
On a congested link, the queue length would just keep increasing up to
around 8000 (for average size packets), at which point the hardware would
start refusing the packets and they would begin to be dropped.
This prevented the qdisc layer from effectively managing and prioritising
packets, as essentially all packets were being allowed into the driver
queue and then were being dropped by the hardware.
This change only restarts the queue if the length is less than 1000
(MAX_OUT_QUEUE_DEPTH).
Reviewed-by: Kyeong Yoo <kyeong.yoo@alliedtelesis.co.nz> Reviewed-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Reviewed-by: Richard Laing <richard.laing@alliedtelesis.co.nz> Signed-off-by: Luuk Paulussen <luuk.paulussen@alliedtelesis.co.nz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the network portion of a frame is preceded by more than 14 bytes of
data, the checksum calculated in the HW is done over the wrong data and
is put in the wrong place.
In our use case an Octeon ethernet controller is connected to a Broadcom
switch chip. Extra data is included in the frame prior to egressing the
Octeon ethernet (i.e. 4 bytes of an 802.1Q tag, 4 bytes of a proprietary
BCM tag later stripped by the switch chip). This extra data causes the
checksum calculation to be incorrect.
The fix in this patch is to make use of the network header offset of the
skb. This enables the checksum to be calculated correctly.
This has been tested in both the configuration with the switch chip in
the egress path (as described above) and in a simple connection direct
to the wire.
Reviewed-by: Richard Laing <richard.laing@alliedtelesis.co.nz> Reviewed-by: Tim Beale <tim.beale@alliedtelesis.co.nz> Reviewed-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Signed-off-by: Hamish Martin <hamish.martin@alliedtelesis.co.nz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The interrupt handler for DPRC IRQs is added. DPRC IRQs are
generated for hot plug events related to DPAA2 objects in a given
DPRC. These events include, creating/destroying DPAA2 objects in
the DPRC, changing the "plugged" state of DPAA2 objects and moving
objects between DPRCs.
Signed-off-by: J. German Rivera <German.Rivera@freescale.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: fsl-mc: Populate the IRQ pool for an MC bus instance
Scan the corresponding DPRC container to get total count
of IRQs needed by all its child DPAA2 objects. Then,
preallocate a set of MSI IRQs with the DPRC's ICID
(GIT-ITS device Id) to populate the the DPRC's IRQ pool.
Each child DPAA2 object in the DPRC and the DPRC object itself
will allocate their necessary MSI IRQs from the DPRC's IRQ pool,
in their driver probe function.
Signed-off-by: J. German Rivera <German.Rivera@freescale.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: fsl-mc: Changed DPRC built-in portal's mc_io to be atomic
The DPRC built-in portal's mc_io is used to send commands to the MC
to program MSIs for MC objects. This is done by the
fsl_mc_msi_write_msg() callback, which is invoked by the generic MSI
layer with interrupts disabled. As a result, the mc_io used in
fsl_mc_msi_write_msg needs to be an atomic mc_io.
Signed-off-by: J. German Rivera <German.Rivera@freescale.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: fsl-mc: Extended MC bus allocator to include IRQs
All the IRQs for DPAA2 objects in the same DPRC must use
the ICID of that DPRC, as their device Id in the GIC-ITS.
Thus, all these IRQs must share the same ITT table in the GIC.
As a result, a pool of IRQs with the same device Id must be
preallocated per DPRC (fsl-mc bus instance). So, the fsl-mc
bus object allocator is extended to also provide services
to allocate IRQs to DPAA2 devices, from their parent fsl-mc bus
IRQ pool.
Signed-off-by: J. German Rivera <German.Rivera@freescale.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: fsl-mc: Added generic MSI support for FSL-MC devices
Created an MSI domain for the fsl-mc bus-- including functions
to create a domain, find a domain, alloc/free domain irqs, and
bus specific overrides for domain and irq_chip ops.
Signed-off-by: J. German Rivera <German.Rivera@freescale.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Alexey Tulia [Sat, 21 Nov 2015 12:36:47 +0000 (15:36 +0300)]
staging: vt6656: fix definitions of DEVICE_FLAGS_* flags
test_bit and set_bit take the bit number to operate on, rather than a
mask. This patch fixes the DEVICE_FLAGS_* definitions so that they represent
the bit index in priv->flags as opposed to the mask returned by the
BIT macro.
rtw_AcceptAddbaReq is a static variable, it is set once and never
modified. It is referenced only once, to assign its value to a member
of struct registry_priv with practically the same name.
Get rid of the variable, and move the meaningful part of the comment
near the declaration of the relevant field of struct registry_priv.
Raises a new checkpatch issue, which is fixed in a later commit:
CHECK: Avoid CamelCase: <bAcceptAddbaReq>
This line is connented since the initial import in commit 7b464c9fa5cc
("staging: r8188eu: Add files for new driver - part 4") and there's no
comment stating how it could ever be useful.
Mike Rapoport [Sun, 17 Jan 2016 18:04:19 +0000 (20:04 +0200)]
staging: sm750fb: change definition of SYSTEM_CTRL multi-bit fields
Use more straight-forward definitions for multi-bit fields of
SYSTEM_CTRL register and replace FIELD_GET/SET for these fields with
open-coded implementation.
Signed-off-by: Mike Rapoport <mike.rapoport@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Mike Rapoport [Sun, 17 Jan 2016 17:59:49 +0000 (19:59 +0200)]
staging: sm750fb: disable PCI device if lynxfb_pci_probe fails
In case of error during lynxfb_pci_probe, the function returned without
calling pci_disable_device. Fix it by adding pci_disable_device on the
error cleanup path.
Signed-off-by: Mike Rapoport <mike.rapoport@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ira Weiny [Tue, 22 Dec 2015 02:57:45 +0000 (21:57 -0500)]
staging/rdma/hfi1: Fix Xmit Wait calculation
Total XMIT wait needs to sum the xmit wait values of all the VLs not just
those requested in the query. Also, make the algorithm used for both
PortStatus and PortDataCounters the same.
Reviewed-by: Arthur Kepner <arthur.kepner@intel.com> Reviewed-by: Breyer, Scott J <scott.j.breyer@intel.com> Signed-off-by: Ira Weiny <iweiny@gmail.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Paul Gortmaker [Tue, 17 Nov 2015 00:58:49 +0000 (19:58 -0500)]
staging: drop kbuild workaround dummy module from top dir
This is essentially a revert of 86f69fe9c069dd8608d238581eea259caa1dfc99
("Staging: workaround build system bug"), but to call it a revert would
imply it was wrong -- rather it seems more like it is just no longer
required anymore.
It doesn't list the full details of the failing use case, but the most
obvious would appear to be a "make allnoconfig" that subsequently had set
just CONFIG_STAGING=y in the .config file, to give:
paul@builder:~/git/linux-head$ grep STAGING ../staging-allno/.config
CONFIG_STAGING=y
# CONFIG_STAGING_MEDIA is not set
# CONFIG_STAGING_BOARD is not set
paul@builder:~/git/linux-head$
After building this .config (on ARM, just to be different), we see that
built-in.o is created, and the final vmlinux links OK:
paul@builder:~/git/linux-head$ ls -l ../staging-allno/drivers/staging/built-in.o
-rw-rw-r-- 1 paul paul 257 Nov 16 18:06 ../staging-allno/drivers/staging/built-in.o
paul@builder:~/git/linux-head$ file ../staging-allno/drivers/staging/built-in.o
../staging-allno/drivers/staging/built-in.o: ELF 32-bit LSB relocatable, ARM, version 1, not stripped
paul@builder:~/git/linux-head$ nm ../staging-allno/drivers/staging/built-in.o
nm: ../staging-allno/drivers/staging/built-in.o: no symbols
paul@builder:~/git/linux-head$ ls -l ../staging-allno/vmlinux
-rwxrwxr-x 1 paul paul 1236326 Nov 16 18:07 ../staging-allno/vmlinux
paul@builder:~/git/linux-head$
I also tested an "allmodconfig" and did not see any problems there
either. Switching back to x86-64 and testing several things there didn't
show any issues either. So it appears we do not need to carry the
workaround in tree any longer.
Cc: linux-kbuild@vger.kernel.org Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> Acked-by: Michal Marek <mmarek@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Pablo G. Gallardo <pggllrd@gmail.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
STAGING: COMEDI: Remove unnecessary typecast of c90 int constant
This patch removes unnecessary typecast of c90 int constant.
Signed-off-by: Pablo G. Gallardo <pggllrd@gmail.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Spencer E. Olson [Wed, 27 Jan 2016 21:28:29 +0000 (14:28 -0700)]
staging: comedi: ni_mio_common: add finite regeneration to dio output
This patch continues the implementation of reinterpreting stop_arg when
stop_src == TRIG_NONE for national instruments cdio output on e/m-series
devices. This is part of a series of patches that allow a user to have a
specific buffer repeated as-is indefinitely. The contents of the DMA
buffer can be left static or changed by the user via mmap access to the DMA
buffer. If the contents are changed by the user, additional munging is not
performed by the driver and only a single call to
comedi_mark_buffer_written should be done. The original behavior is
preserved when stop_arg == 0, as would be the prior use case.
As opposed to analog output, this patch is relatively simple. First, the
digital output capabilities are much more limited/simple as compared to the
analog output device on NI e/m-series hardware, and second, this patch
relies on changes made with the earlier patch to accomplish limiting the
DMA buffer transfer.
Signed-off-by: Spencer E. Olson <olsonse@umich.edu> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Spencer E. Olson [Wed, 27 Jan 2016 21:28:28 +0000 (14:28 -0700)]
staging: comedi: ni_mio_common: adds finite regeneration to AO output
This patch implements for analog output the reinterpretation of stop_arg
when stop_src == TRIG_NONE to allow the user to specify the length of the
buffer that should be repeated. The intent is to allow a user to have a
specific buffer repeated as-is indefinitely. The contents of the DMA
buffer can be left static or changed by the user via mmap access to the DMA
buffer. If the contents are changed by the user, additional munging is not
performed by the driver and only a single call to
comedi_mark_buffer_written should be done.
Signed-off-by: Spencer E. Olson <olsonse@umich.edu> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch implements ni_ao_cmd much more closely organized like NI MHDDK
examples and DAQ-STC pseudo-code. Adds comments with some more specific
references to the DAQ-STC.
For stop_src==TRIG_NONE (continuous output mode of entire buffer), the
count for the UC counter was corrected to represent the maximum count
possible (0xffffff). Prior behavior for stop_src=TRIG_NONE did not
actually follow the DAQ-STC. Furthermore, stop_src==TRIG_NONE now
correctly uses code specialized for either m-series or e-series devices.
It should be noted that stop_src==TRIG_NONE does _not_ with this patch
(or with prior behavior in ni_mio_common) actually implement true
continuous output. Rather, the output is simply configured to operate as a
single buffer output, but where the buffer is as large as is possible with
NI-STC hardware.
Signed-off-by: Spencer E. Olson <olsonse@umich.edu> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Spencer E. Olson [Wed, 27 Jan 2016 21:28:26 +0000 (14:28 -0700)]
staging: comedi: ni_mio_common: make more bits in ao_cmd1 reg be stateful
Bits NISTC_AO_CMD1_DAC0_UPDATE_MODE and NISTC_AO_CMD1_DAC1_UPDATE_MODE are
now saved in the local copy of the AO_CMD1 register. This is more
appropriate than prior methods of setting these bits specifically _both_
in the ni_ao_cmd configuration function _and_ the ni_ao_inttrig trigger
function. With this patch, the bits are only specifically called out now
in the ni_ao_cmd configuration function. In the ni_ao_inttrig trigger
function, only the UI_ARM, UC_ARM, BC_ARM bits of the ao_cmd1 register are
specifically called out. Each of these bits is a strobe bit, while the
DAC[0-1]_UPDATE_MODE bits are simple write bits.
Signed-off-by: Spencer E. Olson <olsonse@umich.edu> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch implements ni_ao_reset much more closely organized like NI MHDDK
examples and DAQ-STC pseudo-code. Adds comments with some more specific
references to the DAQ-STC.
Signed-off-by: Spencer E. Olson <olsonse@umich.edu> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Spencer E. Olson [Wed, 27 Jan 2016 21:28:24 +0000 (14:28 -0700)]
staging: comedi: mite: enable continuous regeneration of finite samples
This change enables the mite DMA controller to only transfer the amount of
data needed by a command. By default, the old behavior of transferring the
entire comedi DMA data buffer is still in effect. These changes allow a
command to only transmit a limited portion of that data buffer as needed.
This patch begins to reinterprets stop_arg when stop_src == TRIG_NONE to
allow the user to specify the length of the buffer that should be repeated.
The intent is to allow a user to have a specific buffer repeated as-is
indefinitely. The contents of the DMA buffer can be left static or changed
by the user via mmap access to the DMA buffer. If the contents are changed
by the user, additional munging is not performed by the driver and only a
single call to comedi_mark_buffer_written should be done.
Signed-off-by: Spencer E. Olson <olsonse@umich.edu> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Spencer E. Olson [Tue, 12 Jan 2016 18:05:10 +0000 (11:05 -0700)]
staging: comedi: ni_mio_common: use CR_CHAN more consistently
Generally, the CR_CHAN macro is/should be used to access the relevant bits
for channel identification in cmd->*_arg when the corresponding
cmd->*_src==TRIG_EXT, including cmd->convert_arg in this case.
This patch does not fix a bug per se, as NISTC_AI_MODE1_CONVERT_SRC() already
masks the value sufficiently, but using CR_CHAN() here makes the code clearer as
it avoids passing some irrelevant bits to NISTC_AI_MODE1_CONVERT_SRC() in the
first place.
Signed-off-by: Spencer E. Olson <olsonse@umich.edu> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / mirror_ubuntu-eoan-kernel.git / log