| Age | Commit message (Collapse) | Author |
|---|
|
Before this fix, the rsc table header file was included directly in C
code. And the table was getting optimized away due to no direct
reference to the resource table. This resulted in failure of test
downloaded through user space loader.
Instead added the configuration to include the resource table (without
vdev and vring) through configuration file, which is more cleaner.
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
Make sure only one instance of the LAD daemon is running
at a time. This implementation is based on syslink's
daemon.
Signed-off-by: Misael Lopez Cruz <misael.lopez@ti.com>
|
|
The LAD daemon was only deleting specific FIFOs for new clients
whose PIDs had a corresponding old FIFO left over. This was not
effective as the client side checks if the FIFO doesn't exist
before sending the LAD_CONNECT command, so the old FIFO was
not deleted.
The cleanup is changed to delete any old FIFO left over from
previous LAD sessions. The cleanup implementation is based on
wpa_supplicant's.
Signed-off-by: Vishal Mahaveer <vishalm@ti.com>
Signed-off-by: Misael Lopez Cruz <misael.lopez@ti.com>
|
|
It is not required that the local endpoint be a certain value
when creating an OmapRpc instance. Since we are performing a
NameMap of the created endpoint to the host, the endpoint can
be any value. In fact, hard-coding the value prevents multiple
MmServiceMgr instances from being created on a processor.
This patch changes the port value from the hard-coded value
of 59 to be a value that is assigned by the RPMessage_create
call. In this way, multiple MmServiceMgr instances can
be created.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
The hwspinlock_user is a character driver that exposes ioctls to
lock and unlock a given hwspinlock instance. The hwspinlock_user
creates the /dev/hwspinlock device file.
The hwspinlock_user driver is used as the first option when available,
otherwise the previous /dev/mem approach is used.
The hwspinlock_user.h uapi header file is kept locally as it may not
be present in all supported kernels.
Signed-off-by: Misael Lopez Cruz <misael.lopez@ti.com>
|
|
The spinlock device is opened in the GateHWSpinlock_start()
function, but it's never closed. The device is now being
closed in GateHWSpinlock_stop().
Signed-off-by: Misael Lopez Cruz <misael.lopez@ti.com>
|
|
The GateMP daemon uses /dev/mem to access SR0, where the gate
usage arrays are kept. /dev/mem may not be present in all
production systems due to the security risk that it implies.
A userspace I/O driver approach can be used as an alternative.
The Linux kernel needs to expose a UIO named 'sr0' in order
to access the shared region from userspace. The daemon is now
capable of finding such UIO entry if exists, and otherwise
fall back to the previous /dev/mem based approach.
Signed-off-by: Misael Lopez Cruz <misael.lopez@ti.com>
|
|
|
|
- These types are no longer used in IPC
- Moving forward the idea is to move to using standard C99 types
instead of defining types here in this local file.
- This also fixes issues with Int8, where other components defining
the same changed it to use int8_t which is defined a signed char
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
- Update to previous patch
- Interrupt configuration for only K2G is changed
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
- old references to ti.bios is getting deprecated in BIOS 6.45 and later
releases. Removing here to avoid build issues with new bios releases.
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
The device name under the BIOS target configuration
is different for DSP and ARM. Settings needs an alias
for the ARM core to build ARM-RTOS even on existing platforms
Signed-off-by: John Godbey <j-godbey@ti.com>
|
|
- The DSP interrupt number and host interrupt numbers are different for
K2G compared to other Keystone devices.
- Seperate entry need created to capture the K2G settings.
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
Changed the formatting of the #ifdef definitions to
#if defined(xdc_target__isaCompatible_xxxx)
to better follow the existing code.
Signed-off-by: John Godbey <j-godbey@ti.com>
|
|
Added the hooks for ARM_RTOS support on the K2E and K2L platforms in Settings.xs.
Signed-off-by: John Godbey <j-godbey@ti.com>
|
|
pthread definitions are updated in Android M, update definitions
accordingly. Using PLATFORM_SDK_VERSION to determine versions < Android M
Android version numbers can be found here:
https://source.android.com/source/build-numbers.html#platform-code-names-versions-api-levels-and-ndk-releases
Signed-off-by: Vishal Mahaveer <vishalm@ti.com>
|
|
Added an ifdef to the file to let ARM-RTOS build cleanly without modifying
other files.
Signed-off-by: John Godbey <j-godbey@ti.com>
|
|
- Added support for the Interrupt module running on ARM-RTOS.
- Updated Interrupt.c to handle the ARM side.
- Added an entry for hardware semaphore support in Settings.xs.
Signed-off-by:John Godbey <j-godbey@ti.com>
|
|
-ti.ipc.family.tci6638.Interrupt is deprecated and any reference to
this module need to removed from test/example code
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
- MultiProc_rprocSetId and LAD_RPROC_SETID are no longer needed
as the out of order download issue is taken care in the
Remoteproc/RPMSG kernel driver
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
- Check in autoreconf-generated files for 66AK2G support.
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
The C66AK2G device is an A15 + C66 DSP, similar in hardware
to the TCI663X family of devices. This commit adds Linux-side
support.
Note that this commit doesn't update the autotools-generated
files. That will be done in a separate commit.
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
The TCI6638/TCI6614 resource table structure definitions are
incorrect - the offset variable was defined to be an array
of 13, whereas the real valid entries are either 2 or 3,
thereby mis-representing the offsets of the resources in the
actual resource table when compared to the resource structure
definition. Fix this properly by adjusting the array length.
While at this, also remove the current TYPE_CARVEOUT entry as
well, Keystone family of SoCs do not have an MMU, so can not
really support dynamically allocated addresses for text or
data sections.
Signed-off-by: Suman Anna <s-anna@ti.com>
|
|
A recent change to the BIOS-side GateMPSupportNull exposed an issue in
the GateMP module in which it may attempt to map an address of 0x0.
Previously, when GateMPSupportNull was used, the number of resources
was set as 1 and a valid address was obtained. With the update,
the number of resources for GateMPSupportNull is now 0. In this case,
the GateMP code will attempt to map an address of 0x0, of size 0,
which will fail.
This patch fixes the code to first check that the number of resources
is not 0, and only then proceed with the mapping. If the number of
resources is 0, then we skip the mapping for that remote gate.
This patch also makes a minor update to the LOGs for the failure
cases for mapping remoteCustom1InUse and remoteCustom2InUse to
print the proper remote InUse array.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
- RPMSG2 supports RPMSG_NS_2_0 which is default for
remoteproc driver in kernel version 4.1
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
- Note this will break backward compatiblity with keystone2 remoteproc driver
based on kernel version 3.10.x
- RPMSG_NS_2_0 is default for keystone2 remoteproc driver
based on kernel version 4.1
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
- Update VirtQueue driver to remove the per core patching of Vring address
- This patch was required with the keystone 2 remoteproc driver
based on kernel version 3.10.x
- No longer needed for kernel drivers added on the 4.1 2015 LTS kernel
for keystone2
- Currently putting under #ifdef in case customer wants use it to use
with older kernel
Signed-off-by: Sam Nelson <sam.nelson@ti.com>
|
|
Attempting to compile the examples and tests in Linux with
the c++ compiler resulted in the following errors:
In file included from MessageQApp.c:46:0:
../../../packages/ti/ipc/Ipc.h:148:22: error: ‘<anonymous>’ has incomplete type
../../../packages/ti/ipc/Ipc.h:148:26: error: invalid use of ‘Void {aka void}’
In file included from ../../../linux/include/_lad.h:50:0,
from ../../../linux/include/ladclient.h:58,
from LAD_client.c:48:
../../../linux/include/_GateMP.h:78:3: error: conflicting declaration 'typedef struct GateMP_Object GateMP_Object'
In file included from ../../../linux/include/_lad.h:49:0,
from ../../../linux/include/ladclient.h:58,
from LAD_client.c:48:
../../../packages/ti/ipc/GateMP.h:222:16: error: 'struct GateMP_Object' has a previous declaration as 'struct GateMP_Object
This patch fixes these compilation issues so that the Linux examples can
be compiled with c++. Note that this patch does not have complete
changes to allow IPC to be compiled using the c++ compiler, but only
the examples.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
Some functions in the List module are called with interrupts disabled. When
these functions write to the slog (IPC_DEBUG_SLOG_LEVEL=7), a hard hang
is occasionally observed. This patch disables the entry/exit trace in these
functions to prevent the system from entering the deadlock state.
Signed-off-by: VW <vwan@ti.com>
|
|
Previously, after recovery, an application using MessageQ is expected to
terminate. We are now improving this by having MessageQ_get/put return
MessageQ_E_SHUTDOWN after recovery is performed, upon which the application
can cleanup, and call Ipc_stop and Ipc_start to "restart" IPC. After that the
application can use IPC again.
This commit also adds a Fault test to validate this behavior.
This is to address CQ SDOCM00116175.
Signed-off-by: VW <vwan@ti.com>
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
|
|
This test was incorrectly using a socket that was connected
to a remote endpoint to receive messages on the HOST. The
remote core was using the source address to send the response
when communicating with an rpmsg-proto channel, which is not
a valid use of the rpmsg-proto channel. Only sockets which
have been bound to a local endpoint by the application should
be used to receive messages.
When the remote core sends a message to an endpoint that
is associated with a channel, there is a kernel crash because
the private data for that endpoint is not as expected.
Update the test case to properly use rpmsg-proto. With these
updates, the test case now passes and there is no kernel
crash.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
During error recovery, the IPC resource manager will dump
the current traces for each remote core to a log file, if
a log file was specified when launching the IPC. At the
beginning of each core's trace dump, a header is printed
to identify which core's traces were being printed. This
trace was incorrectly using "i" instead of "id" to index
the firmware array, resulting in an incorrect remote core
id being printed in the header. This could even index the
array beyond the valid range and cause a crash in IPC.
This patch fixes the index so that it uses the proper
variable ("id"), and also modifies the header to print
the more user-friendly remote proc name instead of the
remote proc ID.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
Consolidate Interrupt and NotifyCircSetup modules from package
ti.ipc.family.tci6638 into package ti.sdo.ipc.family.tci663x.
This package is now able to support both Linux and non-Linux
configurations. Replace Interrupt binding in VirtQueue with a
proxy. Existing configurations will receive a warning when using
the deprecated modules. The Interrupt module snoops for VirtQueue
and NotifyCircSetup to avoid host conflict. Remove deprecated
modules from custom build flow.
|
|
The host program MessageQApp was modified to pass the message Id
in the payload instead of the message header. Update DSP program
with SyncMsg data type to access message Id in payload.
|
|
Add config validate function to ensure host processor is first
in the MultiProc name list array. The Interrupt module makes
this assumption; it will fail otherwise.
|
|
During the recovery, in deinit_ipc(), the IPC resource manager will
try to dump the remote core trace information to a file. There is a
possibility that the trace buffer may have become corrupted as a result
of the crash on the remote core and that the write index, which resides
in the trace buffer, has become corrupted as well. In this situation,
there is a potential that the trace dump code will read beyond the mapped
buffer for the TRACE BUF memory, causing the system to crash and be
unable to recover. At this point a system restart is required to
recover the system.
There should be a check to make sure that the trace dump code is not
indexing beyond the end of the trace buffer.
Signed-off-by: Buddy Liong <buddy.budiono@ti.com>
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
Add new device name 66AK2G to top-level products.mak file. Add
device TCI66AK2G02 alias to ti.sdo.ipc.family.Settings.
|
|
Add documentation to Linux transport interface header files.
|
|
Delete the INetworkTransportDummy interface. This file was
overlooked when removing the transport network hack.
|
|
Replace xdc.loadPackage of ti.sdo.ipcmgr with xdc.useModule
of ti.sdo.ipcmgr.IpcMgr.
|
|
Support has been removed for the f28 device. Remove old references.
|
|
Document the fact that NameServer_get cannot be called from Hwi
or Swi context. This was the main point of the bug report.
|
|
When using the custom IPC build option, the link order of
some packages is pseudo-random. This results in linker errors.
In ti.sdo.ipc.close(), add a package dependency to modules
participating in the custom build flow.
|
|
Delete C28-related packages, and remove build-related support.
Signed-off-by: Chris Ring <cring@ti.com>
|
|
This patch introduces the changes needed so that
the QNX IPC can be compiled against QNX SDP 6.5.0.
Where needed, a check is made to make one call
when compiling against QNX SDP 6.6 and another
call when compiling against QNX SDP 6.5 in order
to take advantage of the newer functionality of
QNX SDP 6.6.
Signed-off-by: Buddy Liong <a0270631@ti.com>
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
Add missing __FAR__ access qualifier to extern symbol declarations.
|
|
Remove the module TransportNetworkDummy from the transports package
and remove reference to the module in MessageQ.
|
|
Add missing __FAR__ access qualifier to extern symbol declarations.
|
|
Fix metaonly module function to return zero instead of one.
|
