OSDN Git Service

android-x86/kernel.git
13 years agodmaengine/dmatest: Pass timeout via module params
Viresh Kumar [Tue, 22 Mar 2011 11:57:25 +0000 (17:27 +0530)]
dmaengine/dmatest: Pass timeout via module params

When we try to test all channels present on our controller together, some
channels of lower priority may be very slow as compared to others. If number of
transfers is unlimited, some channels may timeout and will not finish within 3
seconds. Thus, while doing such regress testing we may need to have higher value
of timeouts. This patch adds support for passing timeout value via module
parameters. Default value is 3 msec, a negative value means max timeout
possible.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma: let IMX_DMA depend on IMX_HAVE_DMA_V1 instead of an explicit list of SoCs
Uwe Kleine-König [Tue, 22 Mar 2011 09:35:17 +0000 (10:35 +0100)]
dma: let IMX_DMA depend on IMX_HAVE_DMA_V1 instead of an explicit list of SoCs

As a side effect this makes IMX_DMA selectable on i.MX21 again, because
the symbol ARCH_MX21 doesn't exist (MACH_MX21 would have been more correct).

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: make halt behave nicely on all supported controllers
Ira Snyder [Thu, 3 Mar 2011 07:55:01 +0000 (07:55 +0000)]
fsldma: make halt behave nicely on all supported controllers

The original dma_halt() function set the CA (channel abort) bit on both
the 83xx and 85xx controllers. This is incorrect on the 83xx, where this
bit means TEM (transfer error mask) instead. The 83xx doesn't support
channel abort, so we only do this operation on 85xx.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: reduce locking during descriptor cleanup
Ira Snyder [Thu, 3 Mar 2011 07:55:00 +0000 (07:55 +0000)]
fsldma: reduce locking during descriptor cleanup

This merges the fsl_chan_ld_cleanup() function into the dma_do_tasklet()
function to reduce locking overhead. In the best case, we will be able
to keep the DMA controller busy while we are freeing used descriptors.
In all cases, the spinlock is grabbed two times fewer than before on
each transaction.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: support async_tx dependencies and automatic unmapping
Ira Snyder [Thu, 3 Mar 2011 07:54:59 +0000 (07:54 +0000)]
fsldma: support async_tx dependencies and automatic unmapping

Previous to this patch, the dma_run_dependencies() function has been
called while holding desc_lock. This function can call tx_submit() for
other descriptors, which may try to re-grab the lock. Avoid this by
moving the descriptors to be cleaned up to a temporary list, and
dropping the lock before cleanup.

At the same time, add support for automatic unmapping of src and dst
buffers, as offered by the DMAEngine API.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: fix controller lockups
Ira Snyder [Thu, 3 Mar 2011 07:54:58 +0000 (07:54 +0000)]
fsldma: fix controller lockups

Enabling poisoning in the dmapool API quickly showed that the DMA
controller was fetching descriptors that should not have been in use.
This has caused intermittent controller lockups during testing.

I have been unable to figure out the exact set of conditions which cause
this to happen. However, I believe it is related to the driver using the
hardware registers to track whether the controller is busy or not. The
code can incorrectly decide that the hardware is idle due to lag between
register writes and the hardware actually becoming busy.

To fix this, the driver has been reworked to explicitly track the state
of the hardware, rather than try to guess what it is doing based on the
register values.

This has passed dmatest with 10 threads per channel, 100000 iterations
per thread several times without error. Previously, this would fail
within a few seconds.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: minor codingstyle and consistency fixes
Ira Snyder [Thu, 3 Mar 2011 07:54:57 +0000 (07:54 +0000)]
fsldma: minor codingstyle and consistency fixes

This fixes some minor violations of the coding style. It also changes
the style of the device_prep_dma_*() function definitions so they are
identical.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: improve link descriptor debugging
Ira Snyder [Thu, 3 Mar 2011 07:54:56 +0000 (07:54 +0000)]
fsldma: improve link descriptor debugging

This adds better tracking to link descriptor allocations, callbacks, and
frees. This makes it much easier to track errors with link descriptors.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: use channel name in printk output
Ira Snyder [Thu, 3 Mar 2011 07:54:55 +0000 (07:54 +0000)]
fsldma: use channel name in printk output

This makes debugging the driver much easier when multiple channels are
running concurrently. In addition, you can see how much descriptor
memory each channel has allocated via the dmapool API in sysfs.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agofsldma: move related helper functions near each other
Ira Snyder [Thu, 3 Mar 2011 07:54:54 +0000 (07:54 +0000)]
fsldma: move related helper functions near each other

This is a purely cosmetic cleanup. It is nice to have related functions
right next to each other in the code.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodmatest: fix automatic buffer unmap type
Ira Snyder [Thu, 3 Mar 2011 07:54:53 +0000 (07:54 +0000)]
dmatest: fix automatic buffer unmap type

The dmatest code relies on the DMAEngine API to automatically call
dma_unmap_single() on src buffers. The flags it passes are incorrect,
fix them.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agoMerge branch 'for_dan' of git://git.infradead.org/users/vkoul/slave-dma into dmaengine
Dan Williams [Fri, 11 Mar 2011 22:56:58 +0000 (14:56 -0800)]
Merge branch 'for_dan' of git://git.infradead.org/users/vkoul/slave-dma into dmaengine

* 'for_dan' of git://git.infradead.org/users/vkoul/slave-dma:
  drivers, pch_dma: Fix warning when CONFIG_PM=n.
  dmaengine/dw_dmac fix: use readl & writel instead of __raw_readl & __raw_writel
  avr32: at32ap700x: Specify DMA Flow Controller, Src and Dst msize
  dw_dmac: Setting Default Burst length for transfers as 16.
  dw_dmac: Allow src/dst msize & flow controller to be configured at runtime
  dw_dmac: Changing type of src_master and dest_master to u8.
  dw_dmac: Pass Channel Priority from platform_data
  dw_dmac: Pass Channel Allocation Order from platform_data
  dw_dmac: Mark all tx_descriptors with DMA_CRTL_ACK after xfer finish
  dw_dmac: Change value of DWC_MAX_COUNT to 4095.
  dw_dmac: Adding support for 64 bit access width for memcpy xfers
  dw_dmac: Calling dwc_scan_descriptors from dwc_tx_status() after taking lock
  dw_dmac: Move single descriptor from dwc->queue to dwc->active_list in dwc_complete_all
  dw_dmac: Replace module_init() with subsys_initcall()
  dw_dmac: Remove compilation dependency from AVR32 and put on HAVE_CLK
  dmaengine: mxs-dma: add dma support for i.MX23/28
  pch_dma: set the number of array correctly
  pch_dma: fix kernel error issue

13 years agoMerge branch 'dmaengine-fixes' into dmaengine
Dan Williams [Fri, 11 Mar 2011 22:45:54 +0000 (14:45 -0800)]
Merge branch 'dmaengine-fixes' into dmaengine

13 years agodrivers, pch_dma: Fix warning when CONFIG_PM=n.
Rakib Mullick [Sun, 6 Mar 2011 11:26:10 +0000 (17:26 +0600)]
drivers, pch_dma: Fix warning when CONFIG_PM=n.

When CONFIG_PM=n, we get the following warning:

drivers/dma/pch_dma.c:741: warning: ‘pch_dma_suspend’ defined but not used
drivers/dma/pch_dma.c:755: warning: ‘pch_dma_resume’ defined but not used

To fix it, wrap pch_dma_{suspend,resume} and
pch_dma_{save,restore}_regs functions with CONFIG_PM.

Signed-off-by: Rakib Mullick <rakib.mullick@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodmaengine/dw_dmac fix: use readl & writel instead of __raw_readl & __raw_writel
Viresh Kumar [Fri, 4 Mar 2011 09:28:32 +0000 (14:58 +0530)]
dmaengine/dw_dmac fix: use readl & writel instead of __raw_readl & __raw_writel

On ARMv7 cores, device memory mapped as Normal Non-cacheable, may not guarantee
ordered access causing failures in device drivers that do not use the mandatory
memory barriers. readl & writel versions contain necessary memory barriers for
this.

commit 79f64dbf68c8a9779a7e9a25e0a9f0217a25b57a: "ARM: 6273/1: Add barriers to
the I/O accessors if ARM_DMA_MEM_BUFFERABLE" can be referred for more
information on this.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agoavr32: at32ap700x: Specify DMA Flow Controller, Src and Dst msize
Viresh Kumar [Thu, 3 Mar 2011 10:17:26 +0000 (15:47 +0530)]
avr32: at32ap700x: Specify DMA Flow Controller, Src and Dst msize

Now that the dw_dmac DMA driver supports configurable Flow Controller, source
and destination burst or msize, we need to specify which ones to use. Msize or
burst size was previously hardcoded to 1, Flow controller was DMA for both
M2P & P2M transfers.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Setting Default Burst length for transfers as 16.
Viresh Kumar [Thu, 3 Mar 2011 10:17:25 +0000 (15:47 +0530)]
dw_dmac: Setting Default Burst length for transfers as 16.

This patch sets default Burst length for all transfer to 16. This will
enhance performance when user doesn't have any chan->private data.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Allow src/dst msize & flow controller to be configured at runtime
Viresh KUMAR [Fri, 4 Mar 2011 10:12:51 +0000 (15:42 +0530)]
dw_dmac: Allow src/dst msize & flow controller to be configured at runtime

Msize or Burst Size is peripheral dependent in case of prep_slave_sg and
cyclic_prep transfers, and in case of memcpy transfers it is platform dependent.
So msize configuration must come from platform data.

Also some peripherals (ex: JPEG), need to be flow controller for dma transfers,
so this information in case of slave_sg & cyclic_prep transfers must come from
platform data.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Changing type of src_master and dest_master to u8.
Viresh Kumar [Thu, 3 Mar 2011 10:17:23 +0000 (15:47 +0530)]
dw_dmac: Changing type of src_master and dest_master to u8.

src_master & dest_master don't required u32 as they have values limited to u8
only. Also their description is missing from doc style comment. This patch
fixes above mentioned issues.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Pass Channel Priority from platform_data
Viresh Kumar [Thu, 3 Mar 2011 10:17:22 +0000 (15:47 +0530)]
dw_dmac: Pass Channel Priority from platform_data

In Synopsys designware, channel priority is programmable. This patch adds
support for passing channel priority through platform data. By default Ascending
channel priority will be followed, i.e. channel 0 will get highest priority and
channel 7 will get lowest.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Pass Channel Allocation Order from platform_data
Viresh Kumar [Thu, 3 Mar 2011 10:17:21 +0000 (15:47 +0530)]
dw_dmac: Pass Channel Allocation Order from platform_data

In SPEAr Platform channels 4-7 have more Fifo depth. So we must get better
channel first. This patch introduces concept of channel allocation order in
dw_dmac. If user doesn't pass anything or 0, than normal (ascending) channel
allocation will follow, else channels will be allocated in descending order.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Mark all tx_descriptors with DMA_CRTL_ACK after xfer finish
Viresh Kumar [Thu, 3 Mar 2011 10:17:20 +0000 (15:47 +0530)]
dw_dmac: Mark all tx_descriptors with DMA_CRTL_ACK after xfer finish

dwc_desc_get checks all descriptors for DMA_CTRL_ACK before allocating them for
transfers. And descriptors are not marked with DMA_CRTL_ACK after transfer
finishes. Thus descriptor once used is not usable again. This patch marks
descriptors with DMA_CRTL_ACK after dma xfer finishes

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Change value of DWC_MAX_COUNT to 4095.
Viresh Kumar [Fri, 4 Mar 2011 10:12:50 +0000 (15:42 +0530)]
dw_dmac: Change value of DWC_MAX_COUNT to 4095.

Every descriptor can transfer a maximum count of 4095 (12 bits, in control reg),
So we must have DWC_MAX_COUNT as 4095 instead of 2048.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Adding support for 64 bit access width for memcpy xfers
Viresh Kumar [Thu, 3 Mar 2011 10:17:18 +0000 (15:47 +0530)]
dw_dmac: Adding support for 64 bit access width for memcpy xfers

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Calling dwc_scan_descriptors from dwc_tx_status() after taking lock
Viresh Kumar [Thu, 3 Mar 2011 10:17:17 +0000 (15:47 +0530)]
dw_dmac: Calling dwc_scan_descriptors from dwc_tx_status() after taking lock

Lock must be taken before calling dwc_scan_descriptors, as this may
access/modify shared data and queues. dwc_tx_status wasn't taking lock before
calling this routine. This patch add code that takes lock before calling
dwc_scan_descriptors.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Move single descriptor from dwc->queue to dwc->active_list in dwc_complete_all
Viresh Kumar [Thu, 3 Mar 2011 10:17:16 +0000 (15:47 +0530)]
dw_dmac: Move single descriptor from dwc->queue to dwc->active_list in dwc_complete_all

dwc_complete_all and other routines was removing all descriptors from dwc->queue
and pushing them to dwc->active_list. Only one was required to be removed. Also
we are calling dwc_dostart, once list is fixed.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Replace module_init() with subsys_initcall()
Viresh Kumar [Thu, 3 Mar 2011 10:17:15 +0000 (15:47 +0530)]
dw_dmac: Replace module_init() with subsys_initcall()

In some cases users of dw_dmac are initialized before dw_dmac, and if they try
to use dw_dmac, they simply fail. So its better we register init() routine
of driver using subsys_initcall() instead of module_init(), so that dma driver
is available at the earliest possible.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodw_dmac: Remove compilation dependency from AVR32 and put on HAVE_CLK
Viresh Kumar [Thu, 3 Mar 2011 10:17:14 +0000 (15:47 +0530)]
dw_dmac: Remove compilation dependency from AVR32 and put on HAVE_CLK

This driver will now be used in atleast two platforms AVR32 & ARM. And there is
no actual hardware dependency of this driver over AVR32 or ARM. So this
dependency can be removed altogether.

Also dw_dmac driver uses clk framework and must have compilation dependency on
HAVE_CLK

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agodmaengine: mxs-dma: add dma support for i.MX23/28
Shawn Guo [Sat, 26 Feb 2011 16:47:42 +0000 (00:47 +0800)]
dmaengine: mxs-dma: add dma support for i.MX23/28

This patch adds dma support for Freescale MXS-based SoC i.MX23/28,
including apbh-dma and apbx-dma.

* apbh-dma and apbx-dma are supported in the driver as two mxs-dma
  instances.

* apbh-dma is different between mx23 and mx28, hardware version
  register is used to differentiate.

* mxs-dma supports pio function besides data transfer.  The driver
  uses dma_data_direction DMA_NONE to identify the pio mode, and
  steals sgl and sg_len to get pio words and numbers from clients.

* mxs dmaengine has some very specific features, like sense function
  and the special NAND support (nand_lock, nand_wait4ready).  These
  are too specific to implemented in generic dmaengine driver.

* The driver refers to imx-sdma and only a single descriptor is
  statically assigned to each channel.

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agopch_dma: set the number of array correctly
Tomoya MORINAGA [Fri, 18 Feb 2011 04:31:21 +0000 (10:01 +0530)]
pch_dma: set the number of array correctly

set the number of array correctly.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agopch_dma: fix kernel error issue
Tomoya MORINAGA [Fri, 18 Feb 2011 04:31:20 +0000 (10:01 +0530)]
pch_dma: fix kernel error issue

fix the following kernel error

------------[ cut here ]------------
WARNING: at kernel/softirq.c:159 _local_bh_enable_ip.clone.5+0x35/0x71()
Hardware name: To be filled by O.E.M.
Modules linked in: pch_uart pch_dma fuse mga drm cpufreq_ondemand acpi_cpufreq mperf ip6t_REJECT nf_conntrack_ipv6 nf_defrag_ipv6 ip6table_filter ip6_tables ipv6 uinput snd_hda_codec_realtek snd_hda_intel snd_hda_codec matroxfb_base snd_hwdep 8250_pnp snd_seq snd_seq_device matroxfb_DAC1064 snd_pcm joydev 8250 matroxfb_accel snd_timer matroxfb_Ti3026 ppdev pegasus parport_pc snd parport matroxfb_g450 g450_pll serial_core video output matroxfb_misc soundcore snd_page_alloc serio_raw pcspkr ext4 jbd2 crc16 sdhci_pci sdhci mmc_core floppy [last unloaded: scsi_wait_scan]
Pid: 0, comm: swapper Not tainted 2.6.37.upstream_check+ #8
Call Trace:
 [<c0433add>] warn_slowpath_common+0x65/0x7a
 [<c043825b>] ? _local_bh_enable_ip.clone.5+0x35/0x71
 [<c0433b01>] warn_slowpath_null+0xf/0x13
 [<c043825b>] _local_bh_enable_ip.clone.5+0x35/0x71
 [<c043829f>] local_bh_enable_ip+0x8/0xa
 [<c06ec471>] _raw_spin_unlock_bh+0x10/0x12
 [<f82b57dd>] pd_prep_slave_sg+0xba/0x200 [pch_dma]
 [<f82f7b7a>] pch_uart_interrupt+0x44d/0x6aa [pch_uart]
 [<c046fa97>] handle_IRQ_event+0x1d/0x9e
 [<c047146f>] handle_fasteoi_irq+0x90/0xc7
 [<c04713df>] ? handle_fasteoi_irq+0x0/0xc7
 <IRQ>  [<c04045af>] ? do_IRQ+0x3e/0x89
 [<c04035a9>] ? common_interrupt+0x29/0x30
 [<c04400d8>] ? sys_getpriority+0x12d/0x1a2
 [<c058bb2b>] ? arch_local_irq_enable+0x5/0xb
 [<c058c740>] ? acpi_idle_enter_bm+0x22a/0x261
 [<c0648b11>] ? cpuidle_idle_call+0x70/0xa1
 [<c0401f44>] ? cpu_idle+0x49/0x6a
 [<c06d9fc4>] ? rest_init+0x58/0x5a
 [<c089e762>] ? start_kernel+0x2d0/0x2d5
 [<c089e0ce>] ? i386_start_kernel+0xce/0xd5

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
13 years agoMerge branch 'dw_dmac' into dmaengine
Dan Williams [Mon, 14 Feb 2011 11:36:51 +0000 (03:36 -0800)]
Merge branch 'dw_dmac' into dmaengine

13 years agoMerge branch 'dma40' into dmaengine
Dan Williams [Mon, 14 Feb 2011 11:36:44 +0000 (03:36 -0800)]
Merge branch 'dma40' into dmaengine

13 years agodmaengine: add slave-dma maintainer
Dan Williams [Mon, 14 Feb 2011 08:42:08 +0000 (00:42 -0800)]
dmaengine: add slave-dma maintainer

Slave-dma has become the predominant usage model for dmaengine and needs
special attention.  Memory-to-memory dma usage cases will continue to be
maintained by Dan.

Cc: Alan Cox <alan@linux.intel.com>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agoMerge branch 'imx' into dmaengine-fixes
Dan Williams [Mon, 14 Feb 2011 10:40:46 +0000 (02:40 -0800)]
Merge branch 'imx' into dmaengine-fixes

13 years agodma: ipu_idmac: do not lose valid received data in the irq handler
Anatolij Gustschin [Mon, 31 Jan 2011 12:22:29 +0000 (13:22 +0100)]
dma: ipu_idmac: do not lose valid received data in the irq handler

Currently when two or more buffers are queued by the camera driver
and so the double buffering is enabled in the idmac, we lose one
frame comming from CSI since the reporting of arrival of the first
frame is deferred by the DMAIC_7_EOF interrupt handler and reporting
of the arrival of the last frame is not done at all. So when requesting
N frames from the image sensor we actually receive N - 1 frames in
user space.

The reason for this behaviour is that the DMAIC_7_EOF interrupt
handler misleadingly assumes that the CUR_BUF flag is pointing to the
buffer used by the IDMAC. Actually it is not the case since the
CUR_BUF flag will be flipped by the FSU when the FSU is sending the
<TASK>_NEW_FRM_RDY signal when new frame data is delivered by the CSI.
When sending this singal, FSU updates the DMA_CUR_BUF and the
DMA_BUFx_RDY flags: the DMA_CUR_BUF is flipped, the DMA_BUFx_RDY
is cleared, indicating that the frame data is beeing written by
the IDMAC to the pointed buffer. DMA_BUFx_RDY is supposed to be
set to the ready state again by the MCU, when it has handled the
received data. DMAIC_7_CUR_BUF flag won't be flipped here by the
IPU, so waiting for this event in the EOF interrupt handler is wrong.
Actually there is no spurious interrupt as described in the comments,
this is the valid DMAIC_7_EOF interrupt indicating reception of the
frame from CSI.

The patch removes code that waits for flipping of the DMAIC_7_CUR_BUF
flag in the DMAIC_7_EOF interrupt handler. As the comment in the
current code denotes, this waiting doesn't help anyway. As a result
of this removal the reporting of the first arrived frame is not
deferred to the time of arrival of the next frame and the drivers
software flag 'ichan->active_buffer' is in sync with DMAIC_7_CUR_BUF
flag, so the reception of all requested frames works.

This has been verified on the hardware which is triggering the
image sensor by the programmable state machine, allowing to
obtain exact number of frames. On this hardware we do not tolerate
losing frames.

This patch also removes resetting the DMA_BUFx_RDY flags of
all channels in ipu_disable_channel() since transfers on other
DMA channels might be triggered by other running tasks and the
buffers should always be ready for data sending or reception.

Signed-off-by: Anatolij Gustschin <agust@denx.de>
Reviewed-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agoMerge branch 'dmaengine-shawn' into dmaengine
Sascha Hauer [Mon, 31 Jan 2011 11:42:51 +0000 (12:42 +0100)]
Merge branch 'dmaengine-shawn' into dmaengine

13 years agoMerge branch 'dmaengine-sdma' into dmaengine
Sascha Hauer [Mon, 31 Jan 2011 11:42:48 +0000 (12:42 +0100)]
Merge branch 'dmaengine-sdma' into dmaengine

13 years agodmaengine: imx-sdma: fix up param for the last BD in sdma_prep_slave_sg()
Shawn Guo [Wed, 19 Jan 2011 21:50:39 +0000 (05:50 +0800)]
dmaengine: imx-sdma: fix up param for the last BD in sdma_prep_slave_sg()

As per the reference manual, bit "L" should be set while bit "C"
should be cleared for the last buffer descriptor in the non-cyclic
chain, so that sdma can stop trying to find the next BD and end
the transfer.

In case of sdma_prep_slave_sg(), BD_LAST needs to be set and BD_CONT
be cleared for the last BD.

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine: imx-sdma: correct sdmac->status in sdma_handle_channel_loop()
Shawn Guo [Wed, 19 Jan 2011 21:50:38 +0000 (05:50 +0800)]
dmaengine: imx-sdma: correct sdmac->status in sdma_handle_channel_loop()

sdma_handle_channel_loop() is the handler of cyclic tx.  One period
success does not really mean the success of the tx.  Instead of
DMA_SUCCESS, DMA_IN_PROGRESS should be the one to tell.

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine: imx-sdma: return sdmac->status in sdma_tx_status()
Shawn Guo [Wed, 19 Jan 2011 21:50:37 +0000 (05:50 +0800)]
dmaengine: imx-sdma: return sdmac->status in sdma_tx_status()

The sdmac->status was designed to reflect the status of the tx,
so simply return it in sdma_tx_status().  Then dma client can call
dma_async_is_tx_complete() to know the status of the tx.

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine: imx-sdma: set sdmac->status to DMA_ERROR in err_out of sdma_prep_slave_sg()
Shawn Guo [Wed, 19 Jan 2011 21:50:36 +0000 (05:50 +0800)]
dmaengine: imx-sdma: set sdmac->status to DMA_ERROR in err_out of sdma_prep_slave_sg()

sdma_prep_dma_cyclic() sets sdmac->status to DMA_ERROR in err_out,
and sdma_prep_slave_sg() needs to do the same.  Otherwise,
sdmac->status stays at DMA_IN_PROGRESS, which will make the function
return immediately next time it gets called.

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine: imx-sdma: remove IMX_DMA_SG_LOOP handling in sdma_prep_slave_sg()
Shawn Guo [Wed, 19 Jan 2011 21:50:35 +0000 (05:50 +0800)]
dmaengine: imx-sdma: remove IMX_DMA_SG_LOOP handling in sdma_prep_slave_sg()

This is a leftover from the time that the driver did not have
sdma_prep_dma_cyclic callback and implemented sound dma as a looped
sg chain.  And it can be removed now.

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX dma: initialize dma capabilities outside channel loop
Sascha Hauer [Mon, 31 Jan 2011 10:35:59 +0000 (11:35 +0100)]
dmaengine i.MX dma: initialize dma capabilities outside channel loop

The capabilities are device specific fields, not channel specific fields.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX DMA: do not initialize chan_id field
Sascha Hauer [Mon, 31 Jan 2011 10:35:44 +0000 (11:35 +0100)]
dmaengine i.MX DMA: do not initialize chan_id field

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX dma: check sg entries for valid addresses and lengths
Sascha Hauer [Wed, 12 Jan 2011 13:13:23 +0000 (14:13 +0100)]
dmaengine i.MX dma: check sg entries for valid addresses and lengths

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX dma: set maximum segment size for our device
Sascha Hauer [Wed, 12 Jan 2011 12:14:37 +0000 (13:14 +0100)]
dmaengine i.MX dma: set maximum segment size for our device

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX SDMA: reserve channel 0 by not registering it
Sascha Hauer [Mon, 31 Jan 2011 09:56:58 +0000 (10:56 +0100)]
dmaengine i.MX SDMA: reserve channel 0 by not registering it

We need channel 0 of the sdma engine for internal purposes. We
accomplished this by calling dma_request_channel() in the probe
function. This does not work when multiple dma engines are
present which is the case when IPU support for i.MX31/35 is
compiled in. So instead of registering channel 0 and reserving
it afterwards simply do not register it in the first place.
With this the dmaengine channel counting does not match sdma
channel counting anymore, so we have to use sdma channel counting
in the driver.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX SDMA: initialize dma capabilities outside channel loop
Sascha Hauer [Mon, 31 Jan 2011 09:21:35 +0000 (10:21 +0100)]
dmaengine i.MX SDMA: initialize dma capabilities outside channel loop

The capabilities are device specific fields, not channel specific fields.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX SDMA: do not initialize chan_id field
Sascha Hauer [Mon, 31 Jan 2011 09:19:53 +0000 (10:19 +0100)]
dmaengine i.MX SDMA: do not initialize chan_id field

This is bogus as the dmaengine core will overwrite this field.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX sdma: check sg entries for valid addresses and lengths
Sascha Hauer [Wed, 12 Jan 2011 12:02:28 +0000 (13:02 +0100)]
dmaengine i.MX sdma: check sg entries for valid addresses and lengths

This patch lets sdma_prep_slave_sg fail if the entries of an
sg list do not start on multiples of the word size or if the
lengths are not multiple of the word size.
Also, catch the previously unhandled DMA_SLAVE_BUSWIDTH_8_BYTES
and DMA_SLAVE_BUSWIDTH_UNDEFINED cases.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodmaengine i.MX sdma: set maximum segment size for our device
Sascha Hauer [Wed, 12 Jan 2011 11:12:31 +0000 (12:12 +0100)]
dmaengine i.MX sdma: set maximum segment size for our device

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agodma40: cyclic xfer support
Rabin Vincent [Tue, 25 Jan 2011 10:18:35 +0000 (11:18 +0100)]
dma40: cyclic xfer support

Support cyclic transfers, which are useful for ALSA drivers.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: stop ongoing transfers in DMA_TERMINATE_ALL
Rabin Vincent [Tue, 25 Jan 2011 10:18:34 +0000 (11:18 +0100)]
dma40: stop ongoing transfers in DMA_TERMINATE_ALL

The current implementation of DMA_TERMINATE_ALL leaves ongoing transfers
running.  Fix it.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: fix DMA_SG capability and channels
Rabin Vincent [Tue, 25 Jan 2011 10:18:33 +0000 (11:18 +0100)]
dma40: fix DMA_SG capability and channels

The DMA_SG cap is enabled on the wrong channel, and the pointers are repeatedly
set incorrectly.  Fix it and combine the ops settings to a common function.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: handle failure to allocate first LCLA
Rabin Vincent [Tue, 25 Jan 2011 10:18:32 +0000 (11:18 +0100)]
dma40: handle failure to allocate first LCLA

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: extract lcla code into separate function
Rabin Vincent [Tue, 25 Jan 2011 10:18:31 +0000 (11:18 +0100)]
dma40: extract lcla code into separate function

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: use flags to reduce parameter count
Rabin Vincent [Tue, 25 Jan 2011 10:18:30 +0000 (11:18 +0100)]
dma40: use flags to reduce parameter count

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: make d40_log_buf_to_lli static
Rabin Vincent [Tue, 25 Jan 2011 10:18:29 +0000 (11:18 +0100)]
dma40: make d40_log_buf_to_lli static

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: unify src/dst addr check
Rabin Vincent [Tue, 25 Jan 2011 10:18:28 +0000 (11:18 +0100)]
dma40: unify src/dst addr check

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: pass the info pointer all the way to reduce argument count
Rabin Vincent [Tue, 25 Jan 2011 10:18:27 +0000 (11:18 +0100)]
dma40: pass the info pointer all the way to reduce argument count

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: unify d40_log_sg_to_lli funcs for mem and slave
Rabin Vincent [Tue, 25 Jan 2011 10:18:26 +0000 (11:18 +0100)]
dma40: unify d40_log_sg_to_lli funcs for mem and slave

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: combine duplicated code in log_sg_to_dev
Rabin Vincent [Tue, 25 Jan 2011 10:18:25 +0000 (11:18 +0100)]
dma40: combine duplicated code in log_sg_to_dev

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: move lli_load to main source file
Rabin Vincent [Tue, 25 Jan 2011 10:18:24 +0000 (11:18 +0100)]
dma40: move lli_load to main source file

These register writes are better placed in the main source file rather than
ll.c.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: combine mem and slave prep_sg functions
Rabin Vincent [Tue, 25 Jan 2011 10:18:23 +0000 (11:18 +0100)]
dma40: combine mem and slave prep_sg functions

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: remove export of stedma40_memcpy_sg
Rabin Vincent [Tue, 25 Jan 2011 10:18:22 +0000 (11:18 +0100)]
dma40: remove export of stedma40_memcpy_sg

The dmaengine framework has the API for this now.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: combine mem and slave sg-to-lli functions
Rabin Vincent [Tue, 25 Jan 2011 10:18:21 +0000 (11:18 +0100)]
dma40: combine mem and slave sg-to-lli functions

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: remove duplicated dev addr code
Rabin Vincent [Tue, 25 Jan 2011 10:18:20 +0000 (11:18 +0100)]
dma40: remove duplicated dev addr code

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: combine duplicated d40_pool_lli_alloc() calls
Rabin Vincent [Tue, 25 Jan 2011 10:18:19 +0000 (11:18 +0100)]
dma40: combine duplicated d40_pool_lli_alloc() calls

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: combine desc init functions
Rabin Vincent [Tue, 25 Jan 2011 10:18:18 +0000 (11:18 +0100)]
dma40: combine desc init functions

The desc init code can be shared between the mem and slave prep routines.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: implement prep_memcpy as a wrapper around memcpy_sg
Rabin Vincent [Tue, 25 Jan 2011 10:18:17 +0000 (11:18 +0100)]
dma40: implement prep_memcpy as a wrapper around memcpy_sg

To simplify the code.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: remove unnecessary casts
Rabin Vincent [Tue, 25 Jan 2011 10:18:16 +0000 (11:18 +0100)]
dma40: remove unnecessary casts

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: fix DMA API usage for LLIs
Rabin Vincent [Tue, 25 Jan 2011 10:18:15 +0000 (11:18 +0100)]
dma40: fix DMA API usage for LLIs

Map and unmap the LLIs and use dma_sync_single_for_device() appropriately
instead of mapping and never unmapping them.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: fix DMA API usage for LCLA
Rabin Vincent [Tue, 25 Jan 2011 10:18:14 +0000 (11:18 +0100)]
dma40: fix DMA API usage for LCLA

Map the buffer once and use dma_sync*() appropriately instead of mapping the
buffer over and over without unmapping it.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: use sg_dma_address() instead of sg_phys()
Rabin Vincent [Tue, 25 Jan 2011 10:18:13 +0000 (11:18 +0100)]
dma40: use sg_dma_address() instead of sg_phys()

The address to use for DMA should be taken from sg_dma_address() and not
sg_phys().

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: remove unnecessary ALIGN()s
Rabin Vincent [Tue, 25 Jan 2011 10:18:12 +0000 (11:18 +0100)]
dma40: remove unnecessary ALIGN()s

ALIGN(x * y, y) == x * y
ALIGN(aligned + x * y, y) == aligned + x * y

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: allow realtime and priority for event lines
Rabin Vincent [Tue, 25 Jan 2011 10:18:11 +0000 (11:18 +0100)]
dma40: allow realtime and priority for event lines

DB8500v2's DMA40 (revision 3) allows setting event lines as high priority and
real time.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: fix comment to refer to SOCs rather than boards
Rabin Vincent [Tue, 25 Jan 2011 10:18:10 +0000 (11:18 +0100)]
dma40: fix comment to refer to SOCs rather than boards

And add DB8500v2 information.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: use helpers for error functions
Rabin Vincent [Tue, 25 Jan 2011 10:18:09 +0000 (11:18 +0100)]
dma40: use helpers for error functions

Almost every use of dev_err in this driver prints the function name.  Abstract
out wrappers to help with this and reduce code duplication.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: use helpers for channel type check
Rabin Vincent [Tue, 25 Jan 2011 10:18:08 +0000 (11:18 +0100)]
dma40: use helpers for channel type check

The somewhat confusing check

d40c->log_num == D40_PHY_CHAN

and its variants are used in several places to check if a channel is logical or
physical.  Use appropriately named helpers to do this to make the code more
readable.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: use helper for channel registers base
Rabin Vincent [Tue, 25 Jan 2011 10:18:07 +0000 (11:18 +0100)]
dma40: use helper for channel registers base

The register offset computation for accessing channel registers is copy/pasted
in several places.  Create a helper function to do it.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: remove "hardware link with previous jobs" code
Rabin Vincent [Tue, 25 Jan 2011 10:18:06 +0000 (11:18 +0100)]
dma40: remove "hardware link with previous jobs" code

This link in hardware with previous jobs code is:

  - unused, no clients using or requiring this feature
  - incomplete, being implemented only for physical channels
  - broken, only working to perform one link

Remove it.  This also allows us to get rid of the channel pause in the
submit_tx() routine.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: ensure event lines get enabled
Rabin Vincent [Tue, 25 Jan 2011 10:18:05 +0000 (11:18 +0100)]
dma40: ensure event lines get enabled

The controller sometimes fails to register the enable of the event line when
both src and dst event lines are used on the same logical channel.  Implement
the recommended software workaround, which is to retry the write until it
works.

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodma40: make init function static
Rabin Vincent [Tue, 25 Jan 2011 10:18:04 +0000 (11:18 +0100)]
dma40: make init function static

Acked-by: Per Forlin <per.forlin@stericsson.com>
Acked-by: Jonas Aaberg <jonas.aberg@stericsson.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agoDMA: PL08x: fix channel pausing to timeout rather than lockup
Russell King - ARM Linux [Thu, 27 Jan 2011 12:37:44 +0000 (12:37 +0000)]
DMA: PL08x: fix channel pausing to timeout rather than lockup

If a transfer is initiated from memory to a peripheral, then data is
fetched and the channel is marked busy.  This busy status persists until
the HALT bit is set and the queued data has been transfered to the
peripheral.  Waiting indefinitely after setting the HALT bit results in
system lockups.  Timeout this operation, and print an error when this
happens.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agoDMA: PL08x: fix infinite wait when terminating transfers
Russell King - ARM Linux [Thu, 27 Jan 2011 12:32:53 +0000 (12:32 +0000)]
DMA: PL08x: fix infinite wait when terminating transfers

If we try to pause a channel when terminating a transfer, we could end
up spinning for it to become inactive indefinitely, and can result in
an uninterruptible wait requiring a reset to recover from.

Terminating a transfer is supposed to take effect immediately, but may
result in data loss.

To make this clear, rename the function to pl08x_terminate_phy_chan().
Also, make sure it is always consistently called - with the spinlock
held and IRQs disabled, and ensure that the TC and ERR interrupt status
is always cleared.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agoavr32: at32ap700x: specify DMA src and dst masters
Jamie Iles [Fri, 21 Jan 2011 14:11:55 +0000 (14:11 +0000)]
avr32: at32ap700x: specify DMA src and dst masters

Now that the dw_dmac DMA driver supports configurable source and
destination masters we need to specify which ones to use. This was
previously hardcoded to 0 and 1 respectively in the driver.

Acked-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
Signed-off-by: Jamie Iles <jamie@jamieiles.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodmaengine/dw_dmac: provide a mechanism to indicate private devices
Jamie Iles [Fri, 21 Jan 2011 14:11:54 +0000 (14:11 +0000)]
dmaengine/dw_dmac: provide a mechanism to indicate private devices

Some platforms (e.g. Picochip PC3XX) have multiple DMA controllers
where some may be used for slave transfers and others for general
purpose memcpy type transfers. Add a .is_private boolean to the
platform data structure so that controllers can be marked as private
so that the DMA_PRIVATE capability will be set for that controller.

Signed-off-by: Jamie Iles <jamie.iles@picochip.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodmaengine/dw_dmac: allow src/dst masters to be configured at runtime
Jamie Iles [Fri, 21 Jan 2011 14:11:53 +0000 (14:11 +0000)]
dmaengine/dw_dmac: allow src/dst masters to be configured at runtime

Some platforms have flexible mastering capabilities and this needs
to be selected at runtime. If the platform has specified private
data in the form of the dw_dma_slave then fetch the source and
destination masters from here. If this isn't present, default to
the previous of 0 and 1.

v2: cleanup whitespace

Acked-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
Signed-off-by: Jamie Iles <jamie.iles@picochip.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodmaengine/dw_dmac: don't scan descriptors if no xfers in progress
Jamie Iles [Fri, 21 Jan 2011 14:11:52 +0000 (14:11 +0000)]
dmaengine/dw_dmac: don't scan descriptors if no xfers in progress

Some hardware (picoChip picoXCell in particular) sometimes has
the block transfer complete bit being set for a channel after the
whole transfer has completed. If we don't have any transfers in the
active list then don't bother to scan the descriptors. This often
happens in normal operation and doesn't require the channel to be
reset.

v2: cleanup whitespace

Signed-off-by: Jamie Iles <jamie.iles@picochip.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodmaengine: imx-sdma: fix inconsistent naming in sdma_assign_cookie()
Shawn Guo [Mon, 17 Jan 2011 14:39:24 +0000 (22:39 +0800)]
dmaengine: imx-sdma: fix inconsistent naming in sdma_assign_cookie()

Variable name sdma and sdmac are consistently used as the pointer to
sdma_engine and sdma_channel respectively throughout the file.  The
patch fixes the inconsistency seen in function sdma_assign_cookie().

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodmaengine: imx-sdma: propagate error in sdma_probe() instead of returning 0
Shawn Guo [Wed, 19 Jan 2011 11:13:06 +0000 (19:13 +0800)]
dmaengine: imx-sdma: propagate error in sdma_probe() instead of returning 0

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agodmaengine i.MX SDMA: Fix firmware loading
Sascha Hauer [Wed, 12 Jan 2011 10:18:14 +0000 (11:18 +0100)]
dmaengine i.MX SDMA: Fix firmware loading

When loading the microcode to the SDMA engine we have to use
the ram_code_start_addr found in the firmware image. The copy
in the sdma engine is not initialized correctly. This is broken
since:
5b28aa3 dmaengine i.MX SDMA: Allow to run without firmware

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
13 years agoLinux 2.6.38-rc2
Linus Torvalds [Sat, 22 Jan 2011 03:01:34 +0000 (19:01 -0800)]
Linux 2.6.38-rc2

13 years agoMerge branch 'media_fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Sat, 22 Jan 2011 00:50:31 +0000 (16:50 -0800)]
Merge branch 'media_fixes' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'media_fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6: (101 commits)
  [media] staging/lirc: fix mem leaks and ptr err usage
  [media] hdpvr: reduce latency of i2c read/write w/recycled buffer
  [media] hdpvr: enable IR part
  [media] rc/mceusb: timeout should be in ns, not us
  [media] v4l2-device: fix 'use-after-freed' oops
  [media] v4l2-dev: don't memset video_device.dev
  [media] zoran: use video_device_alloc instead of kmalloc
  [media] w9966: zero device state after a detach
  [media] v4l: Fix a use-before-set in the control framework
  [media] v4l: Include linux/videodev2.h in media/v4l2-ctrls.h
  [media] DocBook/v4l: update V4L2 revision and update copyright years
  [media] DocBook/v4l: fix validation error in dev-rds.xml
  [media] v4l2-ctrls: queryctrl shouldn't attempt to replace V4L2_CID_PRIVATE_BASE IDs
  [media] v4l2-ctrls: fix missing 'read-only' check
  [media] pvrusb2: Provide more information about IR units to lirc_zilog and ir-kbd-i2c
  [media] ir-kbd-i2c: Add back defaults setting for Zilog Z8's at addr 0x71
  [media] lirc_zilog: Update TODO.lirc_zilog
  [media] lirc_zilog: Add Andy Walls to copyright notice and authors list
  [media] lirc_zilog: Remove useless struct i2c_driver.command function
  [media] lirc_zilog: Remove unneeded tests for existence of the IR Tx function
  ...

13 years agoKEYS: Fix up comments in key management code
David Howells [Thu, 20 Jan 2011 16:38:33 +0000 (16:38 +0000)]
KEYS: Fix up comments in key management code

Fix up comments in the key management code.  No functional changes.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoKEYS: Do some style cleanup in the key management code.
David Howells [Thu, 20 Jan 2011 16:38:27 +0000 (16:38 +0000)]
KEYS: Do some style cleanup in the key management code.

Do a bit of a style clean up in the key management code.  No functional
changes.

Done using:

  perl -p -i -e 's!^/[*]*/\n!!' security/keys/*.c
  perl -p -i -e 's!} /[*] end [a-z0-9_]*[(][)] [*]/\n!}\n!' security/keys/*.c
  sed -i -s -e ": next" -e N -e 's/^\n[}]$/}/' -e t -e P -e 's/^.*\n//' -e "b next" security/keys/*.c

To remove /*****/ lines, remove comments on the closing brace of a
function to name the function and remove blank lines before the closing
brace of a function.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Fri, 21 Jan 2011 21:44:07 +0000 (13:44 -0800)]
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  cifs: fix up CIFSSMBEcho for unaligned access
  cifs: fix unaligned accesses in cifsConvertToUCS
  cifs: clean up unaligned accesses in cifs_unicode.c
  cifs: fix unaligned access in check2ndT2 and coalesce_t2
  cifs: clean up unaligned accesses in validate_t2
  cifs: use get/put_unaligned functions to access ByteCount
  cifs: move time field in cifsInodeInfo
  cifs: TCP_Server_Info diet
  CIFS: Implement cifs_strict_readv (try #4)
  CIFS: Implement cifs_file_strict_mmap (try #2)
  CIFS: Implement cifs_strict_fsync
  CIFS: Make cifsFileInfo_put work with strict cache mode

13 years agoMerge branch 'fixes-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu
Linus Torvalds [Fri, 21 Jan 2011 21:43:21 +0000 (13:43 -0800)]
Merge branch 'fixes-2.6.38' of git://git./linux/kernel/git/tj/percpu

* 'fixes-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu:
  x86,percpu: Move out of place 64 bit ops into X86_64 section

13 years agoMerge branch 'fixes-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
Linus Torvalds [Fri, 21 Jan 2011 21:38:57 +0000 (13:38 -0800)]
Merge branch 'fixes-2.6.38' of git://git./linux/kernel/git/tj/wq

* 'fixes-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
  workqueue: note the nested NOT_RUNNING test in worker_clr_flags() isn't a noop
  workqueue: relax lockdep annotation on flush_work()