Fixed the bug where the L4 peripheral address table was getting
corrupted after calling mmu module's function to update TLB entry.
this resulted in the consecutive runs failure as wrong addresses
gets programmed due to L4 peripheral table corruption.
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>
Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>

In sysmgr_ioctl, the execution of CMD_SYSMGR_STOPCALLBACK does not set the
return value in cargs.api_status.  This caused an error sometimes because the
status was uninitialized.  Now it sets the status to 0 just like for the cases
of CMD_SYSMGR_LOADCALLBACK and CMD_SYSMGR_STARTCALLBACK.  Note that since the
platform callbacks return void, we must set the api_status manually.
Signed-off-by: Simon Que <sque@ti.com>

In proc4430, ducati_destroy() is called from both proc4430_destroy() and
proc4430_stop().  Since ducati_setup() is called from proc4430_start(), the
call to ducati_destroy() in proc4430_destroy() is invalid.  It generates a
crash.
Signed-off-by: Simon Que <sque@ti.com>

This fixes the bug in hw mmu module where the
table entries that has memory segment information
was getting corrupted after first run. This bug
causes the consecutive runs to fail.
This patch also fixes the memory leak associated with
not calling iounmap and fixes a compile warning as one
of the function is unused.
Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>

Failed statements are printed in the IPC module xxx_destroy() functions
only after checking if the status value indicates an error condition. This
is to avoid confusion by the existing trace statement word 'failed'. This
statement was being printed if the xxx_destroy jumps to exit label because
the module's reference count is not yet zero.
Signed-off-by: Suman Anna <s-anna@ti.com>

This patch modifies the memory addresses in sysmgr to be in
line with the new memory map changes required to use the
TLB entries for Ducati.
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>
Signed-off-by: Suman Anna <s-anna@ti.com>

This fixes the 16MB TLB issue.

Signed-off-by: Ramesh Gupta G<grgupta@ti.com>
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>

This patch programs all the required memory regions for Ducati
as TLBs to provide a sw solution for the hw bug in SDC. A total
of 49MB of physical memory is mapped, and the physical address is
assumed to begin at 0x88000000.
Signed-off-by: Suman Anna <s-anna@ti.com>

This patch de-couples SysManager calls from Proc module.
Added new SysManger calls that get invoked from user
space.
Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>

This patch for the notify module fixes the following issues:

	- Random signal of exit event given to the user
	  thread in the read ioctl call.
	- Memory leak because of the absence of
	  iounmap() call.
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>

Added the following changes
	- Removed array that was used to track ducati
	  driver objects.
	- Removed array introduced in syslink 2-0-0-12
	  migration to track notify driver.
	- Introduced a list implementation for tracking
	  notify objects.
	- Removed the index variable introduced in syslink
	  2-0-0-12 migration.
	- Re-organized notify_ducatidrv_create() function
	  call.
	- Added support for multiple creates to be called
	  with same driver name.
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>

This patch adds the changes to Proc start and
Proc stop based on the validation on Zebu.
Proc start function is extended to enable M3
clock and program Ducati MMU. Proc Stop is extended
to put Cores in reset and disable M3 clock, and the
option to disable individual Core and MMU.
Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>

platform_stop_callback now reads the config info for each module, just like
platform_start_callback does.  The problem was that it didn't read it before,
nor did it save the config info from platform_start_callback for each Ducati
core separately.  Thus only the AppM3 config was saved, and then used for SysM3
cleanup.
Signed-off-by: Simon Que <sque@ti.com>

This patch fixes the bug in notify drv read where
the read thread wouldn't get blocked if the event
list is empty. This bug resulted in board freeze
in Zebu enviornment because the read thread was
blockign all other threads from getting scheduled.

Signed-off-by: Hari Kanigeri<h-kanigeri2@ti.com>

Fix the casting compile warning in platform destroy.
Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>

Added probe and remove functions, added platform_device_add,
platform_device_alloc calls and also added few dev_dbgs to
demo the usage of dev_dbg.
Signed-off-by: Imtiaz Ahmed <x0116819@ti.com>

This patch fixes the issues found in..
	<1> notify_ducati.c: we were not unmapping
		"driver_obj->ctrl_ptr". This is done using
		unmap_ducati_virt_mem() implemented now in
		ducatienabler.
	<2> ducatienabler.c: implemented unmap_ducati_virt_mem()
		to iounmap() the "shm_virt_addr" we get calling
		ioremap() in get_ducati_virt_mem().
	<3> ducatienabler.h: added unmap_ducati_virt_mem().
	<4> notify_ducatidriver.h: added unmap_ducati_virt_mem().
Signed-off-by: Jayan John <x00jayan@ti.com>

This patch fixes an issue found in notify_dispatcher.c where we
were not calling iounmap() on "mailbx_hw_config.mbox_linear_addr"
which we had obtained using ioremap() earlier in ntfy_disp_init().
Signed-off-by: Jayan John <x00jayan@ti.com>

This patch fixes an issue found in platform.c where we were not
unlocking "platform_mem_state.gate" in case of a ioremap() failure.
Signed-off-by: Jayan John <x00jayan@ti.com>

This patch fixes the issues found in..
	<1> platform.c:
		- calling platform_mem_unmap() on all addreses
                	on which we have called platform_mem_map().
		- calling proc_mgr_detach for SysM3 and AppM3.
	<2> -proc4430.c:
		- proc4430_detach() fixed to unmap all reqd.
			addresses.
		- cleanup
Signed-off-by: Jayan John <x00jayan@ti.com>

When a heapbuf object is created, there are two memory buffers allocated -- one
for the handle and one for the object itself.  That is in _heapbuf_create.  In
heapbuf_delete, both are freed.  But in heapbuf_close, only the handle is freed.
The object pointer is not freed and caused a memory leak.  This patch adds a
call to free the object.
Signed-off-by: Simon Que <sque@ti.com>

In the function notify_ducatidrv_unregister_event, it finds the event listener
and deletes the list element containing that listener, but did not free the
listener object itself, which was kmalloced in notify_ducatidrv_register_event.
The listener is now being freed.
Signed-off-by: Simon Que <sque@ti.com>

sysmgr_destroy was not cleaning up platform_mem.  As a result, memory allocated
by platform_mem_setup was not freed by platform_mem_destroy.  This has been
fixed.  Also, platform_mem_destroy was incorrectly freeing the allocated gate
handle.  It now frees it correctly.
Signed-off-by: Simon Que <sque@ti.com>

The memory alignment was not done properly in proc4430_map.  It was not
being aligned to the size of all pages needed to cover the reserved
area.  As a result, not all of it was reserved.  This has been fixed and
tested.
Signed-off-by: Simon Que <sque@ti.com>

This is a patch for near-term to enable MM to validate non-Tiler buffers.
This patch  maps the SW DMM region at 0x80000000 and of size 4MB into
the kernel space and also adds a new shared region for it in
platform_setup function. platform_override_config function is updated
to configure the messageq module with 3 heaps.
Signed-off-by: Suman Anna <s-anna@ti.com>

Syslink destroy phase was failing because of incorrect order of
operations in gatepeterson and notify ducati.
Signed-off-by: Simon Que <sque@ti.com>

Error message strings in sysmgr did not have newline characters at end.
They have been added.
Signed-off-by: Simon Que <sque@ti.com>

Changed the order of heapbuf cleanup operations in
platform_stop_callback.  The heapbuf gate handles are deleted after
heapbuf is closed.  Otherwise the heapbuf close function will hang when
attempting to enter the gate.
Signed-off-by: Simon Que <sque@ti.com>

Thi patch has changes to migrate to syslink version 2.0.0.12
Also added some additional checks. Added -EINVAL return code
for invalid arguments.
Signed-off-by: C A Subramaniam <subramaniam.ca@gmail.com>

Patch Introduces additional tracking of notify driver
handles.
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>
Signed-off-by: Ramesh Gupta G <grgupta@ti.com>

This patch incorporates a number of bug fixes from the Syslink
2.00.00.12 drop. The following is the summary of changes.

1. sharedregion:
	- sharedregion_remove() is corrected to remove the
		sharedregion entry from the desired processor table.
2. platform_mem:
	- fixed memory leaks.
3. nameserver_remotenotify:
	- an atomic_t variable is added to maintain reference counting
		and the state of the module.
	- nameserver_get_local searches directly copy the value into the
		sharedmemory request buffer.
	- additional error checking is added.
4. messageq:
	- Number of priority queues definition moved from messageq source
		code to messageq header file, and is now also used by
		the messageq_transportshm module.
	- messageqq_transportshm module state object also keeps track of
		the individual messageq_transportshm instances created.
5. listmp_sharedmemory:
	- listmp_sharedmemory_open return value can come from nameserver,
		and so is checked against only a negative value for failure
	- sharedregion value is added into the nameserver database rather
		than the absolute sharedaddr.
	- minor changes in variable names.
6. heapbuf:
	- additional checks were added for valid sharedregion ptrs.
	- heapbuf_open return value can come from nameserver, and so
		is checked against only a negative value for failure.
	- minor changes in variable names.
7. gatepeterson:
	- minor changes in gatepeterson module. This module is currently
		missing the gatespinlock for interrupt context protection.
	- additional checks were added for valid sharedregion ptrs.
Signed-off-by: Suman Anna <s-anna@ti.com>

This patch provides the following functionality:
	1. Virt_To_Phys - Given a Co-processor's virtual
	address, the pages that are mapped are retured
	in the buffer passed by the user.
	2. Given a Chiron's virtual address, maps to
	Ducati virtual address. This is a hack for now
	to address Tiler memory until the dynmamic memory
	allocation for Tiler is supported.

	This patch also fixed bug in memory unmap function
	using iommu for Super and Large page cases.
Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>

This patch adds support to use multiple ipc heaps.
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>

This patch cleans up the gt_trace code in the rest of the
IPC modules. printks are added instead in the IPC driver
code. The gt.h header file is no longer included in any
IPC source files. This patch has also fixed couple of
checkpatch warnings in sharedregion module.
Signed-off-by: Suman Anna <s-anna@ti.com>

This patch cleaned up all the gt_traces in messageq modules. Duplicate
BUG_ONs have also been removed. WARN_ONs have been added instead in some
places. Return statements have been added in couple of functions for
successful execution scenarions.
Signed-off-by: Suman Anna <s-anna@ti.com>

This patch cleaned up all the gt_traces in listmp_sharedmemory and listmp
modules. Duplicate BUG_ONs have also been removed. return statements have
been added in couple of functions for successful execution scenarions.
Signed-off-by: Suman Anna <s-anna@ti.com>

This patch cleaned up all the gt_traces in messageq_transportshm and heapbuf
modules. Duplicate BUG_ONs have also been removed. return statements have
been added in couple of functions for successful execution scenarions.
Signed-off-by: Suman Anna <s-anna@ti.com>

Removed unsed global variables. Also, cleaned up the
platform_destroy() function.

Note: Provision for setting separate heapIDs are also
introduced. However, the values set are the same. The
HeapIDs and names can be changed once the corresponding
changes are made on the MPU side samples and the Ducati
side.
Signed-off-by: C A Subramaniam <subramaniam.ca@ti.com>

This patch fixes the minor issues found in heapbuf.c. heapbuf_alloc() now returns failure status for all failures. Earlier we had 0 returned in the case of some failures.
Signed-off-by: Jayan John <x00jayan@ti.com>