Merely using dma_alloc_coherent does not ensure that there is no stale
data left in the caches for the allocated DMA buffer (i.e. that the
affected cacheline may still be dirty).
The original code was doing the following (on AArch64, which
translates a 'flush' into a 'clean + invalidate'):
# during initialisation:
1. allocate buffers via memalign
=> buffers may still be modified (cached, dirty)
# during interrupt processing
2. clean + invalidate buffers
=> may commit stale data from a modified cacheline
3. read from buffers
This could lead to garbage info being written to buffers before
reading them during even-processing.
To make the event processing more robust, we use the following sequence
for the cache-maintenance:
# during initialisation:
1. allocate buffers via memalign
2. clean + invalidate buffers
(we only need the 'invalidate' part, but dwc3_flush_cache()
always performs a 'clean + invalidate')
# during interrupt processing
3. read the buffers
(we know these lines are not cached, due to the previous
invalidation and no other code touching them in-between)
4. clean + invalidate buffers
=> writes back any modification we may have made during event
processing and ensures that the lines are not in the cache
the next time we enter interrupt processing
Note that with the original sequence, we observe reproducible
(depending on the cache state: i.e. running dhcp/usb start before will
upset caches to get us around this) issues in the event processing (a
fatal synchronous abort in dwc3_gadget_uboot_handle_interrupt on the
first time interrupt handling is invoked) when running USB mass
storage emulation on our RK3399-Q7 with data-caches on.
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
The dwc3_flush_cache() call was declared and used inconsistently:
* The declaration assumed 'int' for addresses (a potential issue
when running in a LP64 memory model).
* The invocation cast the address to 'long'.
This change ensures that both the declaration and usage of this
function consistently uses 'uintptr_t' for correct behaviour even
when the allocated buffers (to be flushed) reside outside of the
lower 32bits of memory.
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Remove sys_proto.h inclusion which is not used by the driver.
Signed-off-by: Vincent Tinelli <vincent.tinelli@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
BUILD_BUG_* macros have been defined in several headers. It would
be nice to collect them in include/linux/bug.h like Linux.
This commit is cherry-picking useful macros from include/linux/bug.h
of Linux 4.4.
I did not import BUILD_BUG_ON_MSG() because it would not work if it
is used with include/common.h in U-Boot. I'd like to postpone it
until the root cause (the "error()" macro in include/common.h causes
the name conflict with "__attribute__((error()))") is fixed.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
For u-boot dwc3 driver the scatter gather list support has been removed
from original linux code. It is correct, since we try to send one request
at a time.
However, the cleanup left spurious break, which caused early exit from
loop at dwc3_cleanup_done_reqs() function. As a result the dwc3_gadget_giveback()
wasn't called and caused USB Mass Storage to hang.
This commit removes this problem and refactor the code to remove superfluous
do { } while(1) loop.
Test HW: Odroid XU3 (with ./test/ums/ums_gadget_test.sh)
Signed-off-by: Lukasz Majewski <l.majewski@samsung.com>
Commit "drivers/dwc3: add a workaround for too small OUT requests"
sets max packet for OUT requests when transfer is smaller.
Until this change the default maxpacket for non EP0 EPs was 1024. This is
too much, since UMS LBA size is 512B
Signed-off-by: Lukasz Majewski <l.majewski@samsung.com>
It turned out that current dwc3 gadget code is preparing multiple TRBs
for a transfer. Unfortunately, when multiple requests are in the same
queue, only for the last one the LST (last) ctrl bit is set.
Due to that dwc3 HW executes all TRBs up till the one marked as last.
Unfortunately, UMS requires call of ->complete callback after any send TRB.
This is the reason for "hangs" in executing UMS.
This code simplifies this situation and set each TRB's ctrl field bit to be
last (LST bit).
Signed-off-by: Lukasz Majewski <l.majewski@samsung.com>
There is no point in calling dwc3_thread_interrupt() if no event is
pending. There is also no point in flushing event cache in EVERY loop
iteration.
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
DWC3 hangs on OUT requests smaller than maxpacket size,
so HACK the request length to be at least equal to maxpacket size.
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
The BIT() macro is used only in those places, so it is reasonable to
replace it by a constant value.
Signed-off-by: Lukasz Majewski <l.majewski@samsung.com>
dwc3 can do only max packet aligned transfers. So in case request length
is not max packet aligned and is bigger than DWC3_EP0_BOUNCE_SIZE
two chained TRBs is required to handle the transfer.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Reviewed-by: Lukasz Majewski <l.majewski@samsung.com>
In the linux kernel, non cacheable buffers are used. However in uboot
since there are no APIs to allocate non cacheable memory, all
the buffers should be flushed before using it.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Did a bunch of things to get dwc3/gadget.c compile in u-boot without
build errors and warnings
*) Changed the included header files to that used in u-boot.
*) Used dma_alloc_coherent and dma_free_coherent APIs of u-boot
*) removed sg support
*) remove jiffies and used a simple while loop
*) removed irq support and added a function to call these interrupt handler.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Reviewed-by: Lukasz Majewski <l.majewski@samsung.com>
Removed all pm related operations including pm_runtime APIs,
suspend/resume hooks as support for these are not present in u-boot.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Reviewed-by: Lukasz Majewski <l.majewski@samsung.com>
Removed most of the trace_* APIs from dwc3 driver since tracepoints are not
supported in u-boot. Replaced some of the trace_* API with dev_dbg/dev/vdbg.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Review-by: Lukasz Majewski <l.majewski@samsung.com>
Modified the file header to the format that is used in u-boot. Also
included in the header, the commit in linux kernel from which each of
these files are added.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Reviewed-by: Lukasz Majewski <l.majewski@samsung.com>
Added dwc3 folder from linux kernel 3.19-rc1 (97bf6af1f9)
to u-boot. This will be adapted to work with u-boot in the
following patches.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Reviewed-by: Lukasz Majewski <l.majewski@samsung.com>
Philipp Tomsich
Vincent Tinelli
Masahiro Yamada
Michal Simek
Lukasz Majewski
Marek Szyprowski
Kishon Vijay Abraham I