Some A64 boards (SoPine and Pinebook production batch) use LPDDR3 DRAM
chips.
Add support for LPDDR3 DRAM in the DesignWare-like DRAM controller code.
Real LPDDR3 chips' support is not added yet in this commit.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
Allwinner V3s features a DRAM controller like the on in H3, but with a
DDR2 DRAM.
Add support for it.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
Allwinner V3s SoC features a co-packaged DDR2 DRAM chip, which needs its
timing param.
Add support for it.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Acked-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
The DesignWare-like DRAM code used to set the controller defaultly to
single rank mode, which makes it not able to detect the second rank.
Set the default value to dual rank, thus the rank detection code can
work and finally the rank setting will be the correct value.
Currently we know little about the dual-rank on R40, and the usage
of A15 address line seems to be breaking dual-rank support. The only R40
board currently available (Sinovoip Banana Pi M2 Ultra) uses A15 rather
than dual-rank, thus we cannot do research for it. So dual rank detection
is temporarily disabled on R40.
This change is tested on a Orange Pi One (H3, single rank), a Pine64+
2GiB version (A64, single rank) , a Pinebook early prototype with DDR3
(A64, dual rank) and a SoPine with some LPDDR3 patch (A64, dual CS pins
on one chip).
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
DRAM chip varies, and one code cannot satisfy all DRAMs.
Add options to select a timing set.
Currently only DDR3-1333 (the original set) is added into it.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
Some DDR2 DRAM have only four banks, not eight.
Add code to detect this situation.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
Some Allwinner SoCs features a DesignWare-like controller with only 16
bit bus width.
Add support for them.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
The DesignWare DRAM controller used by H3 and newer SoCs use a bit to
identify whether the DRAM is half-width.
As H3 itself come with 32-bit DRAM, the two modes of the bit used to be
named "MCTL_CR_32BIT" and "MCTL_CR_16BIT", but for SoCs with 16-bit DRAM
they're really 8-bit and 16-bit.
Rename the bit's macro, and also rename the variable name in
dram_sun8i_h3.c.
This commit do not add 16-bit DRAM controller support, but the support
will be introduced in next commit.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
Allwinner SoCs after H3 (e.g. A64, H5, R40, V3s) uses a H3-like
DesignWare DRAM controller, which do not have official free DRAM
initialization code, but can use modified dram_sun8i_h3.c.
Add a invisible option for easier DRAM initialization code reuse.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com>
The R40 seems to have a variant of the memory controller found in
the H3 and A64 SoCs. Adapt the code for use on the R40. The changes
are based on released DRAM code and comparing register dumps from
boot0.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The DRAM controller in the Allwinner H5 SoC is again very similar to
the one in the H3 and A64.
Based on the existing socid parameter, add support for this controller
by reusing the bulk of the code and only deviating where needed.
These new bits set or cleared here and there have been mostly found by
looking at DRAM register dumps after using the H5 boot0 and comparing
them to what we set in the code. So for now it's mostly unclear what
those bits actually mean - hence the missing names and comments.
Also add the delay line parameters taken from the boot0 and libdram
disassembly.
Register setup differences between H5 and H3 are courtesy of Jens Kuske.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Fix the output of the DRAM size on AArch64 SPLs.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>
According to Jens disabling the on-die-termination should set bit 5,
not bit 1 in the respective register. Fix this.
Reported-by: Jens Kuske <jenskuske@gmail.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>
The A64 DRAM controller is very similar to the H3 one,
so the code can be reused with some small changes.
This refactoring does not change the code size for the existing H3 part.
[Andre: rework from #ifdefs to using socid parameters in static
functions, minor fixes, merging in fixes from Jens]
Signed-off-by: Jens Kuske <jenskuske@gmail.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>
So far the DRAM driver for the H3 SoC (and apparently boot0/libdram as
well) only applied coarse delay line settings, with one delay value for
all the data lines in each byte lane and one value for the control lines.
Instead of setting the delays for whole bytes only allow setting it for
each individual bit. Also add support for address/command lane delays.
For the purpose of this patch the rules for the existing coarse settings
were just applied to the new scheme, so the actual register writes don't
change for the H3. Other SoCs will utilize this feature later properly.
With a stock GCC 5.3.0 this increases the dram_sun8i_h3.o code size from
2296 to 2344 Bytes.
[Andre: move delay parameters into macros to ease later sharing, use
defines for numbers of delay registers, extend commit message]
Signed-off-by: Jens Kuske <jenskuske@gmail.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>
The IOCR registers got renamed to BDLR to match the public
documentation of similar controllers.
Signed-off-by: Jens Kuske <jenskuske@gmail.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>
So far the MBUS priority setup was done by writing "magic" values taken
from a DRAM controller register dump after a boot0 run.
By peeking at the Linux (sic!) MBUS driver [1] from the Allwinner BSP
kernel, we learned more about the actual meaning of those bits.
Add macros and refactor the setup function to make the MBUS setup much
more readable and meaningful.
The actual values used now are a transformation of the values used
before, which are assembled by the new code to result in the same register
writes. So this rework does not change any settings, also the code size
stays the same.
The respective source files in the BSP kernel had a proper GPL header,
so lifting this code and information into U-Boot is legal.
[Andre: provide a convenience macro to fit definitions on one line]
[1] https://github.com/longsleep/linux-pine64/blob/lichee-dev-v3.10.65/drivers/bus/sunxi_mbus.c
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>
H3 seems to have a silicon bug breaking the impedance calibration.
This is currently worked around in software by multiple steps
combining the results to replace the wrong values.
Revision A chips need a different workaround, which is present in
the vendor bootloader too, but got overlooked in lack of
information and affected boards till now.
This commit adds a simplified version without correction factor,
which would be 1.00 for all known boards anyway.
Signed-off-by: Jens Kuske <jenskuske@gmail.com>
Reviewed-by: Jagan Teki <jteki@openedev.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Some of the code in arch/arm/cpu/armv7/sunxi is actually armv7 specific, while
most of it is just generic code that could as well be used on an AArch64 SoC.
Move all files that are not really tied to armv7 into a new mach-sunxi
directory.
Signed-off-by: Alexander Graf <agraf@suse.de>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
The read delays were set incorrectly, leading to reliability
issues at higher DRAM clock speeds. This commit adjusts this
to match the vendor boot0 behaviour.
Signed-off-by: Jens Kuske <jenskuske@gmail.com>
Tested-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Based on existing A23/A33 code and the original H3 boot0.
Signed-off-by: Jens Kuske <jenskuske@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Icenowy Zheng
Chen-Yu Tsai
Andre Przywara
Jens Kuske
Philipp Tomsich
Alexander Graf