If we check an index against array bounds, we should do so before
accessing the array and not afterwards.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Move all checks before cache flush and IOC setup.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Add missing cache cleanup before cache disable:
* Flush and invalidate L1 D$ before disabling. Flush and invalidate
SLC before L1 D$ disabling (as it will be bypassed for data)
Otherwise we can lose some data when we disable L1 D$ if this data
isn't flushed to next level cache. Or we can get wrong data if L1 D$
has some entries after enable which we modified when the L1 D$ was
disabled.
* Invalidate L1 I$ before disabling. Otherwise we can execute wrong
instructions after L1 I$ enable if we modified any code when
L1 I$ was disabled.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Add additional cache configuration checks and note about
supported configurations.
It is unlikely to face some configuration in real life but
it's better to be prepared and refuse to work on those.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Implement specialized function to clenup caches (and therefore
sync instruction and data caches) which can be used for cleanup before linux
launch or to sync caches during U-Boot self-relocation.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
If L1 D$ is disabled SLC is bypassed for data and all
load/store requests are sent directly to main memory.
If L1 I$ is disabled SLC is NOT bypassed for instructions
and all instruction requests are fetched through SLC.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Implement icache_enabled() and dcache_enabled() as separate functions
which can be used with "inline" attribute. This is a preparation to
make them always_inline.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Move check for PAE existence into slc_upper_region_init()
instead of its caller as more appropriate place.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
There is a problem with current implementation if we start U-Boot
from ROM, as we use global variables before ther initialization,
so these variables get overwritten when we copy .data section
from ROM.
Instead we move these global variables into our "global data"
structure so that we may really start from ROM.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
There is a problem with current implementation if we start U-Boot
from ROM, as we use global variables before ther initialization,
so these variables get overwritten when we copy .data section
from ROM.
Instead we'll use icache_exists(), dcache_exists(), slc_exists(), pae_exists()
functions which directly check BCRs every time.
In U-Boot case ops are used only during self-relocation and DMA
so we shouldn't be hit by noticeable performance degradation.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
As of today we check SLC status before each call of __slc_rgn_op()
or __slc_entire_op(). So move status check into __slc_rgn_op()
and __slc_entire_op().
As we need to check status before *each* function execution and we
call slc_entire_op() and slc_rgn_op() from different places we add
this check directly into SLC entire/line functions instead of
their callers to avoid code duplication.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Use is_isa_arcv2() function where it is possible instead of
CONFIG_ISA_ARCV2 define check to make code cleaner at the same time
keeping pretty much the same functionality - code in branches
under "if (is_isa_arcv2())" won't be compiled if CONFIG_ISA_ARCV2
is not defined, still we need a couple of CONFIG_ISA_ARCV2
ifdefs to make compiler happy. That's because code in
!is_isa_x() branch gets compiled and only then gets optimized
away.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
As we are planning to get rid of dozens of ifdef's in cache.c we
would better check D$ status before each entire/line operation
then check CONFIG_SYS_DCACHE_OFF config option.
This makes the code cleaner as well as D$ entire/line functions
remain functional even if we enable or disable D$ in run-time.
As we need to check status before *each* function execution and we
call D$ entire/line functions from different places we add
this check directly into D$ entire/line functions instead of
their callers to avoid code duplication.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
We're starting to use more and more BCRs and having their
definitions in-lined in sources becomes a bit annoying
so we move it all to a separate header.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Move IOC initialization from cache_init() to a separate function.
This is the preparation for the next patch where we'll switch
to is_isa_arcv2() function usage instead of "CONFIG_ISA_ARCV2"
ifdef.
Also it makes cache_init function a bit cleaner.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
We don't implement separate flush_dcache_all() intentionally as
entire data cache invalidation is dangerous operation even if we flush
data cache right before invalidation.
There is the real example:
We may get stuck in the following code if we store any context (like
BLINK register) on stack in invalidate_dcache_all() function.
BLINK register is the register where return address is automatically saved
when we do function call with instructions like 'bl'.
void flush_dcache_all() {
__dc_entire_op(OP_FLUSH);
// Other code //
}
void invalidate_dcache_all() {
__dc_entire_op(OP_INV);
// Other code //
}
void foo(void) {
flush_dcache_all();
invalidate_dcache_all();
}
Now let's see what really happens during that code execution:
foo()
|->> call flush_dcache_all
[return address is saved to BLINK register]
[push BLINK] (save to stack) ![point 1]
|->> call __dc_entire_op(OP_FLUSH)
[return address is saved to BLINK register]
[flush L1 D$]
return [jump to BLINK]
<<------
[other flush_dcache_all code]
[pop BLINK] (get from stack)
return [jump to BLINK]
<<------
|->> call invalidate_dcache_all
[return address is saved to BLINK register]
[push BLINK] (save to stack) ![point 2]
|->> call __dc_entire_op(OP_FLUSH)
[return address is saved to BLINK register]
[invalidate L1 D$] ![point 3]
// Oops!!!
// We lose return address from invalidate_dcache_all function:
// we save it to stack and invalidate L1 D$ after that!
return [jump to BLINK]
<<------
[other invalidate_dcache_all code]
[pop BLINK] (get from stack)
// we don't have this data in L1 dcache as we invalidated it in [point 3]
// so we get it from next memory level (for example DDR memory)
// but in the memory we have value which we save in [point 1], which
// is return address from flush_dcache_all function (instead of
// address from current invalidate_dcache_all function which we
// saved in [point 2] !)
return [jump to BLINK]
<<------
// As BLINK points to invalidate_dcache_all, we call it again and
// loop forever.
Fortunately we may do flush and invalidation of D$ with a single one
instruction which automatically mitigates a situation described above.
And because invalidate_dcache_all() isn't used in common U-Boot code we
implement "flush and invalidate dcache all" instead.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Introduce is_isa_arcv2() and is_isa_arcompact() functions.
These functions only check configuration options and return
compile-time constant so they can be used instead of #ifdef's to
to write cleaner code.
Now we can write:
-------------->8---------------
if (is_isa_arcv2())
ioc_configure();
-------------->8---------------
instead of:
-------------->8---------------
ifdef CONFIG_ISA_ARCV2
ioc_configure();
endif
-------------->8---------------
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
As of today __dc_line_op() and __dc_entire_op() support
only separate flush (OP_FLUSH) and invalidate (OP_INV) operations.
Add support of combined flush and invalidate (OP_FLUSH_N_INV)
operation which we planing to use in subsequent patches.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
__cache_line_loop() function was copied from Linux kernel
where per-line instruction cache operations are really used.
In U-Boot we use only entire I$ ops, so we can drop support of
per-line I$ ops from __cache_line_loop() because __cache_line_loop()
is never called with OP_INV_IC parameter.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Move instruction cache entire operation to a separate function
because we are planing to use it in other places like
sync_icache_dcache_all().
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
We improve on 2 things:
1. Only ARC HS family has "dmb" instructions so do compile-time
check for automatically defined macro __ARCHS__.
Previous check for ARCv2 ISA was not good enough because ARC EM
family is v2 ISA as well but still "dmb" instaruction is not
supported in EM family.
2. Still if there's no dedicated instruction for memory barrier
let's at least insert compile-time barrier to make sure
compiler deosn't reorder critical memory operations.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Disabling relocation might be useful on ARC for 2 reasons:
a) For advanced debugging with Synopsys proprietary MetaWare debugger
which is capable of accessing much more specific hardware resources
compared to gdb. For example it may show contents of L1 and L2 caches,
internal states of some hardware blocks etc.
But on the downside MetaWare debugger still cannot work with PIE.
Even though that limitation could be work-arounded with change of ELF's
header and stripping down all debug info but with it we won't have
debug info for source-level debugging which is quite inconvenient.
b) Some platforms which might benefit from usage of U-Boot basically
don't have enough RAM to accommodate relocation of U-Boot so we
keep code in flash and use as much of RAM as possible for more
interesting things.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Bin Meng <bmeng.cn@gmail.com>
Cc: Heiko Schocher <hs@denx.de>
Cc: York Sun <york.sun@nxp.com>
Cc: Stefan Roese <sr@denx.de>
Finally GCC's garbage collector works on ARC so let's use it.
That's what I may see for HSDK:
Before:
text data bss dec hex filename
290153 10068 222616 522837 7fa55 u-boot
After:
text data bss dec hex filename
261999 9460 222360 493819 788fb u-boot
Overall ~5% of memory footprint saved.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
This commit basically reverts two commits:
1. cf628f772e ("arc: arcv1: Disable master/slave check")
2. 6cba327bd9 ("arcv2: Halt non-master cores")
With mentioned commits in-place we experience more trouble than
benefits. In case of SMP Linux kernel this is really required as
we have all the cores running from the very beginning and then we
need to allow master core to do some preparatory work while slaves
are not getting in the way.
In case of U-Boot we:
a) Don't really run more than 1 core in parallel
b) We may use whatever core for that
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
U-Boot is a bit special piese of software because it is being
only executed once on power-on as compared to operating system
for example. That's why we don't care much about performance
optimizations instead we're more concerned about size. And up-to-date
compilers might produce much smaller code compared to
performance-optimized routines copy-pasted from the Linux kernel.
Here's an example:
------------------------------->8--------------------------
--- size_asm_strings.txt
+++ size_c_strings.txt
@@ -1,2 +1,2 @@
text data bss dec hex filename
- 121260 3784 3308 128352 1f560 u-boot
+ 120448 3784 3308 127540 1f234 u-boot
------------------------------->8--------------------------
See we were able to shave off ~800 bytes of .text section.
Also usage of string routines implemented in C gives us an ability
to support more HW flavors for free: generated instructions will match
our target as long as correct compiler option is used.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Fix NOR, SD and QSPI fallback option in case of secure boot failure.
Signed-off-by: Vinitha V Pillai <vinitha.pillai@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Álvaro Fernández Rojas
Heinrich Schuchardt
Eugeniy Paltsev
Alexey Brodkin
Alexey Brodkin
Tom Rini
Vinitha V Pillai