This moves claiming / releasing the interface into the respective
"open" / "close" functions. The USB code in main() is now trimmed
down to:
feldev_init();
handle = open_fel_device(...);
feldev_done(handle);
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
We move USB endpoint detection into the feldev_claim() routine, so
higher level code is no longer involved with that. Also make use of
the "detached" flag within feldev_handle, instead of relying on an
isolated variable.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
This enables us to move forward to a cleaner implementation,
where the "core" fel.c code will become independent of direct
libusb usage. After moving USB code to a separate module,
in the end the libusb handle could become an 'opaque' field of
feldev_handle.
The "device" handle might also be extended later, to provide
(FEL) version data and SoC-specific information (chip ID, SRAM
info, human-readable name).
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
See pull requests #59 and #83.
It looks like various BSD derivates have adopted the more common naming
conventions for the endian conversion macros, which means that we would
attempt to redefine them without need (potentially even to 'historic'
names that no longer exist).
Try to avoid this by properly checking for existing functions first.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
The SID block in the A80 is at 0x01c0e000, with the e-fuses we care
about at offset 0x200 within the block.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
The R40 is marketed as the successor to the A20. The SRAM layout is the
same as the A20, but there doesn't seem to be a secure SRAM block.
The SID block is at a completely different address. The layout is the
same as the newer SoCs, with the e-fuses at an offset of 0x200.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
These functions solve the problem that large readl/writel
transfers might be limited by insufficient (scratch) buffer
size. To solve this, chunks of no more than LCODE_MAX_WORDS
get transferred individually.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
This patch reduces on FEL protocol overhead for the 'multiple'
readl/writel transfers (functions that do word-aligned memory
access on the SoC). The ARM "scratch" code now takes a word count
and is able to work with buffered data, so the host is no longer
required to transfer single words in a piecemeal fashion.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
Besides having fewer lines of code, the #define macros should
also prevent users from accidentally using these names without
braces (i.e. as function pointers). Instead, this will cause
compiler errors now.
soc_info.c: add "A10s" label in comment for SoC ID 0x1625.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
While at it, modify the former "sram_info" identifiers
to carry a broader "soc_info" meaning.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
Nowadays the term "hacking" has gained increasinly negative
connotations. We don't want people to get a wrong impression
and/or fancy ideas, so remove it.
While at it, also get rid of the emphasis on A10.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
Otherwise we could run into ambiguities with "long long",
and Windows compilation is likely to bail out as it might
not understand "%ll".
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
This patch adds shell scripts that deliberately go through some
extra program invocations, e.g. erroneous use of fex2bin. The goal
of these test cases is to improve on code (branch) coverage.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
After all .fex files can be successfully compiled, let's also
test the opposite direction by decompiling the resulting .bin.
The output then has to pass an automated diff against the source
.fex (transformed via some preprocessing with "unify-fex").
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
All tests should go into the new "tests" subdirectory. The idea is
that the separate Makefile in that directory will get invoked via
a top-level "make check".
The tests/Makefile should then take care of running all available
tests, returning an appropriate exit status. Future tests may be
functional, examine code metrics (coverage analysis), or both.
For a start, I'd simply like to check that sunxi-fexc is able to
properly compile all the .fex files from linux-sunxi/sunxi-boards.
(Note: This currently FAILS and will probably require adjustments
to both sunxi-tools and the .fex repository. To work around this,
for now I'm applying patches to fix sunxi-boards.)
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
The previous timeout of 60 seconds was mostly based on scenarios
where large ("write") transfers take place. But it could easily
become annoying if users are awaiting completion of simpler
commands like "read" or "hexdump", and for some reason FEL fails
to respond.
Therefore I've decided to lower the timeout value to 10 seconds,
adjust the maximum chunk size accordingly and - while at it -
improve the source comments documenting their relationship.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
Vendor-provided .fex files have repeatedly shown key-value pairs
(assignments) where the value is an identifier-style string *not*
surrounded by double quotes. The corresponding .bin files confirm
that such values end up as "string" type entries.
So far, our .fex parser has choked on these values. The commit
changes this behavior and treats them as strings now, to allow
processing 'original' .fex without adjustments.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
Vendor-provided .bin files have repeatedly demonstrated that our
previous interpretation of this field as version[0] is likely
wrong. Instead, it seems to represent the file size (in bytes)
of the .bin file.
This commit fixes both decompilation (and header checks) and
generation of .bin files, where it will now store the size to
this field.
TODO: It's unclear whether the 'filesize' needs some specific
alignment (and the .bin corresponding padding). A value of
34864 (0x8830) has already been observed, so any possible
alignment is expected not to exceed 16 bytes (0x10). (Currently
our .bin generator doesn't care about any specific alignment.)
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
This removes the "-g -O0" default (to leave them up to the user
CFLAGS), and adds a switch to ignore "unused result" warnings.
The latter is relevant when trying to compile nand-part.c with
optimizations enabled.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
The previous commit had introduced a new build rule that made
sunxi-pio always (cross-)compile as part of TARGET_TOOLS. This
originated from a misunderstanding, and is wrong - sunxi-pio
should be part of TOOLS instead.
sunxi-pio is a "dual mode" utility. When run natively on a sunxi
SoC, it can mmap() and manipulate the PIO registers directly. But
it also supports file-based operation, to be used in conjunction
with sunxi-fel (after uploading fel-pio.bin thunk code). This
should work over USB from non-sunxi hosts, and thus puts sunxi-pio
in the TOOLS category. See the fel-gpio shell script for details.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
For now, this is for informational purposes (program version
output), but it might also serve as an anchor point for a
"bugfix" release after some changes made to the build system.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
For Linux build testing, we want all targets compiled. To do so
without a cross-toolchain, simply use the host compiler for the
target tools. This can be achieved by setting CROSS_COMPILE to
an empty string.
OSX can't handle this: It neither supports "-static", nor would
it successfully compile meminfo.c. Thus we keep the default
"make", which should only build the 'tools' target.
On Linux we'll also test the install-* targets by requesting
installation to /tmp/sunxi-tools as a post-build step.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
This appends sunxi-meminfo to the TARGET_TOOLS, and adds a new rule
to fix the compilation of sunxi-pio (by making it *cross-compile*
for the target).
Additionally adds a new build target "make install-misc".
For more details, see github issues #69 and #70.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
Fixing the (currently erroneous) compilation of sunxi-pio will
cause "make target-tools" to require a suitable cross-compiler
installed. Otherwise "make target-tools" fails to build.
As that is part of our default target ("make all"), we might
possibly introduce a build breakage on quite a few systems.
Avoid this situation by redefining "make tools" as the default,
and change "make install" to "make install-tools", i.e. limit
the standard targets to those builds that only rely on the host
toolchain. From now, if you actually want to include the cross-
compiling steps, use "make all" or "make install-all" instead.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
Also: Add the winsock2 library to LIBS for Windows. When not
linking against it, the usage of WS2 conversion functions from
portable_endian.h would cause unresolved symbols.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
For Windows portable_endian.h relies on and includes <winsock2.h>.
Thus it needs to be requested first, otherwise other includes might
pull in <windows.h> and cause a preprocessor warning / compilation
failure (observed with MinGW).
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
By defining NO_MMAP it's now possible to avoid the usage of
mmap() and munmap(). This benefits platforms that don't support
these functions, e.g. Windows.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
