Scrub memory content if ECC is enabled and it is not
from warm reset boot.
Enable icache and dcache before scrub memory
and use "DC ZVA" instruction to clear memory
to zeros. This instruction writes a cache line
at a time and it can prevent false ECC error
trigger if write cache line partially.
Signed-off-by: Ley Foon Tan <ley.foon.tan@intel.com>
Stratix 10 maps dram from 0 to 128GB. There is a 2GB hole
in the memory for peripherals and other IO from 2GB to 4GB.
However the dram controller ignores upper address bits for
smaller dram configurations. Example: a 4GB dram
maps to multiple locations, every 4GB on the address.
Signed-off-by: Dalon Westergreen <dalon.westergreen@intel.com>
Signed-off-by: Ley Foon Tan <ley.foon.tan@intel.com>
Move SDRAM size check to SDRAM driver. sdram_calculate_size()
is called in SDRAM initialization already, avoid calling
twice in size check function.
Signed-off-by: Ley Foon Tan <ley.foon.tan@intel.com>
To clean up reset handling for socfpga gen5, port the DDR driver to DM
using UCLASS_RAM and implement proper reset handling.
This gets us rid of one ad-hoc call to socfpga_per_reset().
The gen5 driver is implemented in 2 distinct files. One of it (containing
the calibration training) is not touched much and is kept at using
hard coded addresses since the code grows even more otherwise.
SPL is changed from calling hard into the DDR driver code to just
probing UCLASS_RESET and UCLASS_RAM. It is happy after finding a RAM
driver after that.
Signed-off-by: Simon Goldschmidt <simon.k.r.goldschmidt@gmail.com>
Marvell's switch chips with integrated CPUs (collectively referred to as
MSYS) share common ancestry with the Armada SoCs. Some of the IP blocks
(e.g. xor) are located at different addresses and DFX server exists as a
separate target on the MBUS (on Armada-38x it's just part of the core
complex registers).
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
Inactive chip-selects will give invalid values for read_sample so don't
consider them when trying to determine the overall min/max read sample.
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
[https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell/pull/18]
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
Based on the JEDEC standard JESD79-3F. The tRAS timings should include
the highest speed bins at a given frequency. This is similar to commit
683c67b ("mv_ddr: ddr3: fix tfaw timimg parameter") where the wrong
comparison was used in the initial implementation.
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
[https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell/pull/15]
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
Replace the current rather convoluted code using ad-hoc polling
mechanism with a more straightforward code. Use wait_for_bit_le32()
to poll the DDRCALSTAT register instead of local reimplementation.
It makes no sense to pull for 5 seconds before giving up and trying
to restart the EMIF, so instead wait 500 mSec for the calibration to
complete and if this fails, restart the EMIF and try again. Perform
this 32 times instead of 3 times as the original code did.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Chin Liang See <chin.liang.see@intel.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
Cc: Simon Goldschmidt <simon.k.r.goldschmidt@gmail.com>
Cc: Tien Fong Chee <tien.fong.chee@intel.com>
The EMIF reset code can well use wait_for_bit_le32() instead of all that
convoluted polling code. Reduce the timeout from 100 seconds to 1 second,
since if the EMIF fails to reset itself in 1 second, it's unlikely longer
wait would help. Make sure to clear the EMIF reset request even if the
SEQ2CORE_INT_RESP_BIT isn't asserted.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Chin Liang See <chin.liang.see@intel.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
Cc: Simon Goldschmidt <simon.k.r.goldschmidt@gmail.com>
Cc: Tien Fong Chee <tien.fong.chee@intel.com>
The current EMIF clear timeout handling code was applying bitwise
operations to signed data types and as it was, was extremely hard
to read. Replace it with simple wait_for_bit(). Expand the error
handling to make it more readable too.
This patch also changes the timeout for emif_clear() from 14 hours
to 1 second.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Chin Liang See <chin.liang.see@intel.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
Cc: Simon Goldschmidt <simon.k.r.goldschmidt@gmail.com>
Cc: Tien Fong Chee <tien.fong.chee@intel.com>
The debug print is missing a newline, add it.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Chin Liang See <chin.liang.see@intel.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
Cc: Simon Goldschmidt <simon.k.r.goldschmidt@gmail.com>
Cc: Tien Fong Chee <tien.fong.chee@intel.com>
The Altera Arria10 DDR driver was using constants in a few places
instead of reading registers associated with those constants, fix
this.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Chin Liang See <chin.liang.see@intel.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
Cc: Simon Goldschmidt <simon.k.r.goldschmidt@gmail.com>
Cc: Tien Fong Chee <tien.fong.chee@intel.com>
fsl_ddr_board_options is generally defined in board
board's ddr.c, but some boards like lx2160ardb board
does not need this function.
Defining fsl_ddr_board_options as weak function to
resolve compilation errors for such boards.
Signed-off-by: Priyanka Jain <priyanka.jain@nxp.com>
[PK: Fix checkpatch warnings]
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Add support for lx2160a SoC
-----BEGIN PGP SIGNATURE-----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=aT7d
-----END PGP SIGNATURE-----
Merge tag 'fsl-qoriq-for-v2019.01-rc2' of git://git.denx.de/u-boot-fsl-qoriq
Add TFA boot flow for some Layerscape platforms
Add support for lx2160a SoC
[trini: Add a bunch of missing MAINTAINERS entries]
Signed-off-by: Tom Rini <trini@konsulko.com>
This syncs drivers/ddr/marvell/a38x/ with the mv_ddr-armada-18.09 branch
of https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell.git.
Specifically this syncs with commit 99d772547314 ("Bump mv_ddr to
release armada-18.09.2").
The complete log of changes is best obtained from the mv-ddr-marvell.git
repository but some relevant highlights are:
ddr3: add missing txsdll parameter
ddr3: fix tfaw timimg parameter
ddr3: fix trrd timimg parameter
merge ddr3 topology header file with mv_ddr_topology one
mv_ddr: a38x: fix zero memory size scrubbing issue
The upstream code is incorporated omitting the portions not relevant to
Armada-38x and DDR3. After that a semi-automated step is used to drop
unused features with unifdef
find drivers/ddr/marvell/a38x/ -name '*.[ch]' | \
xargs unifdef -m -UMV_DDR -UMV_DDR_ATF -UCONFIG_DDR4 \
-UCONFIG_APN806 -UCONFIG_MC_STATIC \
-UCONFIG_MC_STATIC_PRINT -UCONFIG_PHY_STATIC \
-UCONFIG_64BIT -UCONFIG_A3700 -UA3900 -UA80X0 \
-UA70X0
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Stefan Roese <sr@denx.de>
Incorrect type of size variable results in 0 being
returned for sdram sizes greater than or equal to
4GB.
Signed-off-by: Dalon Westergreen <dwesterg@gmail.com>
Pass an empty buffer instead of NULL if the hwconfig environment
variable isn't set.
Signed-off-by: Jeremy Gebben <jgebben@sweptlaser.com>
Cc: Stefano Babic <sbabic@denx.de>
Cc: York Sun <york.sun@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
The SDRAM must first be rewritten by zeroes if ECC is used to initialize
the ECC metadata. Make the CPU overwrite the DRAM with zeroes in such a
case. This scrubbing implementation turns the caches on temporarily, then
overwrites the whole RAM with zeroes, flushes the caches and turns them
off again. This provides satisfactory performance.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Chin Liang See <chin.liang.see@intel.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
This function was never used in SPL and the default implementation of
dram_bank_mmu_setup() does the same thing. The only difference is the
part which configures OCRAM as cachable, which doesn't really work as
it covers more than the OCRAM.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Chin Liang See <chin.liang.see@intel.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
This patch enables DDR Kconfig support for Arria 10.
Signed-off-by: Tien Fong Chee <tien.fong.chee@intel.com>
Reviewed-by: Dinh Nguyen <dinguyen@kernel.org>
Current sdram driver is only applied to gen5 device, hence it is better
to rename sdram driver to more specific name which is related to gen5
device.
Signed-off-by: Tien Fong Chee <tien.fong.chee@intel.com>
mv_ddr_build_message.c is generated in Marvell's standalone mv_ddr code.
When imported into u-boot we need to add the appropriate SPDX tag and
re-format it slightly.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Make ddr3_calc_mem_cs_size() global scope and use it in
ddr3_new_tip_ecc_scrub to correctly initialize all of DDR memory.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
This restores support for configuring the timing mode based on the
ddr_topology. This was originally implemented in commit 90bcc3d38d
("driver/ddr: Add support for setting timing in hws_topology_map") but
was removed as part of the upstream sync.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
This syncs drivers/ddr/marvell/a38x/ with the mv_ddr-armada-17.10 branch
of https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell.git.
The upstream code is incorporated omitting the ddr4 and apn806 and
folding the nested a38x directory up one level. After that a
semi-automated step is used to drop unused features with unifdef
find drivers/ddr/marvell/a38x/ -name '*.[ch]' | \
xargs unifdef -m -UMV_DDR -UMV_DDR_ATF -UCONFIG_DDR4 \
-UCONFIG_APN806 -UCONFIG_MC_STATIC \
-UCONFIG_MC_STATIC_PRINT -UCONFIG_PHY_STATIC \
-UCONFIG_64BIT
INTER_REGS_BASE is updated to be defined as SOC_REGS_PHY_BASE.
Some now empty files are removed and the ternary license is replaced
with a SPDX GPL-2.0+ identifier.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
No in-tree code defines SUPPORT_STATIC_DUNIT_CONFIG or
STATIC_ALGO_SUPPORT. Remove ddr3_a38x_mc_static.h and use unifdef to
remove unused sections in the rest of the ddr/marvell/a38x code.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
Move sys_env_device_rev_get() from the ddr training code to
sys_env_lib.c (which currently resides with the serdes code). This
brings sys_env_device_rev_get() into line with sys_env_device_id_get()
and sys_env_model_get().
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
PEX_CFG_DIRECT_ACCESS was defined in ddr3_hws_hw_training_def.h despite
only being used in the serdes code. Move this definition to ctrl_pex.h
where all the other PEX defines are. Also remove the duplicate
definition of PEX_DEVICE_AND_VENDOR_ID which is already defined in
ctrl_pex.h.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
When U-Boot started using SPDX tags we were among the early adopters and
there weren't a lot of other examples to borrow from. So we picked the
area of the file that usually had a full license text and replaced it
with an appropriate SPDX-License-Identifier: entry. Since then, the
Linux Kernel has adopted SPDX tags and they place it as the very first
line in a file (except where shebangs are used, then it's second line)
and with slightly different comment styles than us.
In part due to community overlap, in part due to better tag visibility
and in part for other minor reasons, switch over to that style.
This commit changes all instances where we have a single declared
license in the tag as both the before and after are identical in tag
contents. There's also a few places where I found we did not have a tag
and have introduced one.
Signed-off-by: Tom Rini <trini@konsulko.com>
We have a large number of places where while we historically referenced
gd in the code we no longer do, as well as cases where the code added
that line "just in case" during development and never dropped it.
Signed-off-by: Tom Rini <trini@konsulko.com>
To make this driver easier to be reused, dual-license DDR driver.
Signed-off-by: York Sun <york.sun@nxp.com>
CC: Simon Glass <sjg@chromium.org>
CC: Tom Rini <trini@konsulko.com>
CC: Heinrich Schuchardt <xypron.glpk@gmx.de>
CC: Thomas Schaefer <thomas.schaefer@kontron.com>
CC: Masahiro Yamada <yamada.masahiro@socionext.com>
CC: Robert P. J. Day <rpjday@crashcourse.ca>
CC: Alexander Merkle <alexander.merkle@lauterbach.com>
CC: Joakim Tjernlund <joakim.tjernlund@transmode.se>
CC: Curt Brune <curt@cumulusnetworks.com>
CC: Valentin Longchamp <valentin.longchamp@keymile.com>
CC: Wolfgang Denk <wd@denx.de>
CC: Anatolij Gustschin <agust@denx.de>
CC: Ira W. Snyder <iws@ovro.caltech.edu>
CC: Marek Vasut <marek.vasut@gmail.com>
CC: Kyle Moffett <Kyle.D.Moffett@boeing.com>
CC: Sebastien Carlier <sebastien.carlier@gmail.com>
CC: Stefan Roese <sr@denx.de>
CC: Peter Tyser <ptyser@xes-inc.com>
CC: Paul Gortmaker <paul.gortmaker@windriver.com>
CC: Peter Tyser <ptyser@xes-inc.com>
CC: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
DDR controllers always use binding register to determine the memory
space to perform data initialization. In case of controller interleaving,
the space is doubled, resulting twice long wait. It wasn't too bad until
the memory capacity increases. To reduce the wait time, reduce the
binding space to half and restore it after data initialization.
Three-way interleaving is no longer used and is removed.
Signed-off-by: York Sun <york.sun@nxp.com>
DDR4 RDIMM has some information in SPD to be used to calculate the
control words for register chip. The rest can be found from JEDEC
spec DDR4RCD02.
Signed-off-by: York Sun <york.sun@nxp.com>
On top of RDIMM support, add new register calculation to support
3DS RDIMMs. Only symmetrical 3DS is supported at this time.
Signed-off-by: York Sun <york.sun@nxp.com>
For DDR4, command/address delay in mode registers and parity latency
in timing config register are only needed for UDIMMs, but not RDIMMs.
Add additional register rcw_3 for DDR4 RDIMM. Fix mirrored bit for
dual rank RDIMMs. Set sdram_cfg_3[DIS_MRS_PAR] for RDIMMs. Fix
calculation of timing config registers. Use hexadecimal format for
printing RCW (register control word) registers.
Signed-off-by: York Sun <york.sun@nxp.com>
This driver has been using printf() including filename since it was
added. Convert to using debug() instead.
Signed-off-by: Simon Goldschmidt <sgoldschmidt@de.pepperl-fuchs.com>
wait_for_bit callers use the 32 bit LE version
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>
