'edma_info' structure inside the edma driver represents
a single instance of edma channel controller. Call it
'edma_cc' instead. This also avoids readers confusing
it with an instance of edma_soc_info structre which
carries the platform data for a single channel controller
instance.
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Currently, the ISR in the EDMA driver clears the pending interrupt for all
channels without regard to whether that channel has a registered callback
or not.
This causes problems for devices like DM355/DM365 where the multimedia
accelerator uses EDMA by polling on the interrupt pending bits of some of the
EDMA channels. Since these channels are actually allocated through the Linux
EDMA driver (by an out-of-kernel module), the same shadow region is used by
Linux and accelerator. There a race between the Linux ISR and the polling code
running on the accelerator on the IPR (interrupt pending register).
This patch fixes the issue by making the ISR clear the interrupts only for
those channels which have interrupt enabled. The channels which are allocated
for the purpose of being polled on by the accelerator will not have a callback
function provided and so will not have IER (interrupt enable register) bits set.
Tested on DM365 and OMAP-L137/L138 with audio and MMC/SD (as EDMA users).
Signed-off-by: Anuj Aggarwal <anuj.aggarwal@ti.com>
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
CC: Archith John Bency <archith@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
This patch fixes an issue where a DMA channel can erroneously process an
event generated by a previous transfer. A failure case is where DMA is
being used for SPI transmit and receive channels on OMAP L138. In this
case there is a single bit that controls all event generation from the
SPI peripheral. Therefore it is possible that between when edma_stop()
has been called for the transmit channel on a previous transfer and
edma_start() is called for the transmit channel on a subsequent transfer,
that a transmit event has been generated.
The fix is to clear events in edma_start(). This prevents false events
from being processed when events are enabled for that channel.
Signed-off-by: Brian Niebuhr <bniebuhr@efjohnson.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Currently, the edma_noevent list is passed from platform data.
But on some architectures, there will be many EDMA channels
which will not be used at all. This patch scans all the
platform devices and then builds a list of events which are
not being used. The unused event list will be used to allocate
EDMA channels in case of EDMA_CHANNEL_ANY usage instead of the
edma_noevent being used earlier for this purpose.
This patch is based on David Brownells's suggestion at
http://article.gmane.org/gmane.linux.davinci/15176.
Signed-off-by: Sudhakar Rajashekhara <sudhakar.raj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Though edma_alloc_channel api was looping through the available
channel controllers in EDMA_CHANNEL_ANY case, it was never
returning the channel for 2nd channel controller, if 1st
channel controller had no free channels. This issue has
been fixed with this patch.
Signed-off-by: Sudhakar Rajashekhara <sudhakar.raj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Some architectures have only one channel controller, but the
edma_alloc_channel api loops twice to findout the free channel
available in EDMA_CHANNEL_ANY case. A new variable has been
introduced to keep count of number of channel controllers being
used on a particular architecture.
Signed-off-by: Sudhakar Rajashekhara <sudhakar.raj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Currently, edma_alloc_channel api is returning the channel
number without prepending the controller on which the
channel was allocated. So, if a channel is allocated on
2nd controller, calls subsequent to edma_alloc_channel would
never know that channel was allocated on the 2nd controller,
and continue to operate on 1st controller, resulting in edma
failure. This patch fixes this issue.
Signed-off-by: Sudhakar Rajashekhara <sudhakar.raj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
There have accumulated quite a lot of them after the code reorganizations...
In several cases I had to replace #include <linux/dma-mapping.h> which wasn't
needed directly but happened to #include <linux/err.h> which was needed.
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
The patch itself does not change the functionality of
any existing code. PARAM entries in the present GIT kernel
are referred to as slots. New API's being added to the
EDMA driver were referring to these PARAM entries as
"params". This patch updates the terminolgy used in the
EDMA driver.
Signed-off-by: Sandeep Paulraj <s-paulraj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
The reserve_contiguous_params function is used to reserve
a set of contiguous PARAMs. If we do not find a complete
set of contiguous PARAMs, the functions still has to free
every PARAM that it found to be free in the process of finding a
complete set and thus marked as "in use".
This patch mainly deals with correctly handling the
freeing of PARAMs.
Signed-off-by: Sandeep Paulraj <s-paulraj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
The edma_alloc_cont_slots API is used for obtaining a set of
contiguous slots. When we use the "_ANY" option with this
API, by definition of this option it is suppossed to start
looking for a set of contiguous slots starting from slot 64 for
DaVinci SOC's and 32 for DA8xx SOC's. This has been explained in
the API description in the driver itself. So when we use the
"_ANY" option with this API, the slot number passed as
an argument should be a "don't care".
This patch takes care of this condition mentioned above.
When checking to see if the starting slot is a valid number,
it checks to make sure that the "_ANY" option is not used.
Signed-off-by: Sandeep Paulraj <s-paulraj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
In the edma_free_cont_slots API, the variable slot was being modified
and then used in the for loop.
This results in incorrect behaviour when the API is used.
Signed-off-by: Sandeep Paulraj <s-paulraj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
For best performance, codecs often setup linked triggered
transfers with a contiguous block of params, and that is when
this API is used. Setup/configuration of these parameter RAMs
is most efficient if they are contiguous.
There is an API to allocate a set of contiguous parameter RAMs and
a corresponding API to free a set of contiguous parameter RAMs
Signed-off-by: Sandeep Paulraj <s-paulraj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
EDMA queues in DM365 are a little different than those
on other DaVinci's. On DM365 Q0 and Q1 have the larger
FIFO size. We want Q0 and Q1 to be used by codecs and
DVSDK demos.
MMC driver is the only driver which uses the flag
'EVENTQ_DEFAULT'. So MMC driver should be using Q2 instead of
Q1 on DM365.
This patch allows us to declare a "default queue" from
SOC specific code. If it is not declared then the EDMA
driver assumes a default of queue 1.
Signed-off-by: Sandeep Paulraj <s-paulraj@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
- restructure to support multiple channel controllers by using
additional struct resources for each CC
- interface changes visible to EDMA clients
Introduce macros to build IDs from controller and channel number,
and to extract them. Modify the edma_alloc_slot function to take an
extra argument for the controller.
Also update ASoC drivers to use API. ASoC changes
Acked-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
- Move queue related mappings to dm<soc>.c
EDMA in DM355 and DM644x has two transfer controllers while DM646x
has four transfer controllers. Moving the queue to tc mapping and
queue priority mapping to dm<soc>.c will be helpful to probe these
mappings from platform device so that the machine_is_* testing will
be avoided.
- add channel mapping logic
Channel mapping logic is introduced in dm646x EDMA. This implies
that there is no fixed association for a channel number to a
parameter entry number. In other words, using the DMA channel
mapping registers (DCHMAPn), a PaRAM entry can be mapped to any
channel. While in the case of dm644x and dm355 there is a fixed
mapping between the EDMA channel and Param entry number.
Signed-off-by: Naresh Medisetty <naresh@ti.com>
Signed-off-by: Sudhakar Rajashekhara <sudhakar.raj@ti.com>
Reviewed-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Original code for 2.6.10 and 2.6.28 series done by Texas Instruments
and MontaVista, but major updates and rework done by Troy Kisky and
David Brownell.
Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Sudhakar Rajashekhara <sudhakar.raj@ti.com>
Cc: Troy Kisky <troy.kisky@boundarydevices.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>