When using only a single DDR chip select only assert M_ODT[0] on write.
Do not assert it on read and do not assert M_ODT[1] at all. Also set
tODT_OFF_WR to 0x9 which contradicts the recommendation from the
functional spec but is what Marvell's binary training blob does and
seems to give better results when ODT is active during writes.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
We are now using an env_ prefix for environment functions. Rename these
two functions for consistency. Also add function comments in common.h.
Quite a few places use getenv() in a condition context, provoking a
warning from checkpatch. These are fixed up in this patch also.
Suggested-by: Wolfgang Denk <wd@denx.de>
Signed-off-by: Simon Glass <sjg@chromium.org>
The DDR3 training code for Marvell A38X currently computes 1t timing
when given board topology map of the Turris Omnia, but Omnia needs 2t.
This patch adds support for enforcing the 2t timing in struct
hws_topology_map, through a new enum hws_timing, which can assume
following values:
HWS_TIM_DEFAULT - default behaviour, compute whether to enable 2t
from the number of CSs
HWS_TIM_1T - enforce 1t
HWS_TIM_2T - enforce 2t
This patch also sets all the board topology maps (db-88f6820-amc,
db-88f6820-gp, controlcenterdc and clearfog) to have timing set to
HWS_TIM_DEFAULT.
Signed-off-by: Marek Behun <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
We should not have an arch-specific header file in common.h. Adjust the
board files a little so it is not needed, and drop it.
Signed-off-by: Simon Glass <sjg@chromium.org>
The declarations should not be in common.h. Move them to the arch-specific
headers.
Signed-off-by: Simon Glass <sjg@chromium.org>
[trini: Fixup thinko defined(FSL_LSCH3) -> defined(CONFIG_FSL_LSCH3)]
Signed-off-by: Tom Rini <trini@konsulko.com>
(pdimm[0].data_width >= 32) || (pdimm[0].data_width <= 40)
is always true.
We should use && here.
The problem was indicated by cppcheck.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Depending on DDR configuration, gcc-6.x will show up unused-const-
variable messages. Use __maybe_unused specifier for all dynamic_odt
variable definitions to remove these warnings.
Memory footprint will not increase as gcc will optimize out unused
constants.
Signed-off-by: Thomas Schaefer <thomas.schaefer@kontron.com>
Signed-off-by: York Sun <york.sun@nxp.com>
Convert Altera DDR SDRAM driver to use Kconfig method.
Enable ALTERA_SDRAM by default if it is on Gen5 target.
Arria 10 will have different driver.
Signed-off-by: Tien Fong Chee <tien.fong.chee@intel.com>
Signed-off-by: Ley Foon Tan <ley.foon.tan@intel.com>
This allows us to use the same DRAM init function on all archs. Add a
dummy function for arc, which does not use DRAM init here.
Signed-off-by: Simon Glass <sjg@chromium.org>
[trini: Dummy function on nios2]
Signed-off-by: Tom Rini <trini@konsulko.com>
At present we cannot use this function as an init sequence call without a
wrapper, since it returns the RAM size. Adjust it to set the RAM size in
global_data instead, and return 0 on success.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
These two macros are used for the same thing, the total number of DDR
controllers for a given SoC. Use SYS_NUM_DDR_CTRLS in Kconfig and
merge existing usage.
Signed-off-by: York Sun <york.sun@nxp.com>
Use option NUM_DDR_CONTROLLERS in ddr Kconfig and clean up existing
usage in ls102xa and fsl-layerscape. Remove all powerpc macros in
config header and board header files.
Signed-off-by: York Sun <york.sun@nxp.com>
Create driver/ddr/fsl/Kconfig and move existing options. Clean up
existing macros.
Signed-off-by: York Sun <york.sun@nxp.com>
[trini: Migrate sbc8641d, xpedite537x and MPC8536DS, run a moveconfig.py -s]
Signed-off-by: Tom Rini <trini@konsulko.com>
- add additional function erratum_a009942_check_cpo to check if the
board needs tuning CPO calibration for optimal setting.
- move ERRATUM_A009942(with revision to check cpo_sample option) from
fsl_ddr_gen4.c to ctrl_regs.c for reuse on all DDR4/DDR3 parts.
- move ERRATUM_A008378 from fsl_ddr_gen4.c to ctrl_regs.c
- remove obsolete ERRATUM_A004934 which is replaced with ERRATUM_A009942.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
[YS: Replaced CONFIG_QEMU_E500 with CONFIG_ARCH_QEMU_E500]
Reviewed-by: York Sun <york.sun@nxp.com>
Fix following warning in case multiple erratum macro was not defined.
warning: unused variable 'tmp'
warning: unused variable 'ddr_freq'
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
To enable configuration of sdr.ctrlcfg.extratime1 register which enable
extra clocks for read to write command timing. This is critical to
ensure successful LPDDR2 interface
Signed-off-by: Chin Liang See <clsee@altera.com>
Cc: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Fix various misspellings of:
* deprecated
* partition
* preceding,preceded
* preparation
* its versus it's
* export
* existing
* scenario
* redundant
* remaining
* value
* architecture
Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Reviewed-by: Jagan Teki <jteki@openedev.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Instead of using multiple macros, a data structure is used to pass
board-specific parameters to MMDC DDR driver.
Signed-off-by: York Sun <york.sun@nxp.com>
CC: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
When CONFIG_SYS_FSL_ERRATUM_A009801 is defined but
CONFIG_SYS_FSL_ERRATUM_A008511 not defined, there is compile error
that temp32 undeclared, this patch fixes it.
Signed-off-by: Shaohui Xie <Shaohui.Xie@nxp.com>
Signed-off-by: Gong Qianyu <Qianyu.Gong@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
This general MMDC driver adds basic support for Freescale MMDC
(Multi Mode DDR Controller). Currently MMDC is integrated on ARMv8
LS1012A SoC for DDR3L, there will be a update to this driver to
support more flexible configuration if new features (DDR4, multiple
controllers/chip selections, etc) are implimented in future.
Meantime, reuse common MMDC driver for LS1012ARDB/LS1012AQDS/
LS1012AFRDM.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
DDR controller 5.2.1 has this erratum A008511 partially fixed.
The workaround needs to be adjusted to take advantage of Vref
training. This patch enables the training and force output
enable to be off.
Erratum A009803 requires the controller to be idel before enabling
address parity. It was combined with workaround for A008511. With
new A008511 flow, this flow needs to be changed to enabling
data init (D_INIT) after the address parity is enabled.
Signed-off-by: York Sun <york.sun@nxp.com>
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
32 more debug registers are added for newer DDR controllers.
Signed-off-by: York Sun <york.sun@nxp.com>
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Commit 5605dc6 tried to fix wr_lat bit in timing_cfg_2, but the
change was wrong. wr_lat has 5 bits with MSB at [13] and lower
4 bits at [9:12], in big-endian convention.
Signed-off-by: York Sun <york.sun@nxp.com>
Reported-by: Thomas Schaefer <Thomas.Schaefer@kontron.com>
Add condition of checking the enabled of address parity
for erratum A-009803, if parity is not enabled, the
workaround of erratum A-009803 should not be applied.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Commit 34e026f9 added one extra bit to wr_lat for timing_cfg_2, but
with wrong bit position. It is bit 13 in big-endian, or left shift
18 from LSB. This error hasn't had any impact because we don't have
fast enough DDR4 using the extra bit so far.
Signed-off-by: York Sun <york.sun@nxp.com>
The clk_adjust is of SDRAM_CLK_CNTL[5:8] 4-bits on MPC85xx and P-series,
but is of SDRAM_CLK_CNTL[5:9] 5-bits on T-series and LS-series SoCs.
We should update it to adapt the case that clk_adjust is odd data.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
For reason unknown, recently, the DDR init code writers are really fond
of hiding some small floating point operating deep in their creations.
This patch removes one from the Marvell A38x code.
Instead of returning size of chip as float from ddr3_get_device_size()
in GiB units, return it as int in MiB units. Since this would interfere
with the huge switch code in ddr3_calc_mem_cs_size(), rework the code
to match the change.
Before this patch, the cs_mem_size variable could have these values:
( { 16, 32 } x { 8, 16 } x { 0.01, 0.5, 1, 2, 4, 8 } ) / 8 =
{ 0.000000, 0.001250, 0.002500, 0.005000, 0.062500, 0.125000,
0.250000, 0.500000, 1.000000, 2.000000, 4.000000, }
The switch code checked for a subset of the resulting RAM sizes, which
is in range 128 MiB ... 2048 MiB.
With this patch, the cs_mem_size variable can have these values:
( { 16, 32 } x { 8, 16 } x { 0, 512, 1024, 2048, 4096, 8192 } ) / 8 =
{ 0, 64, 128, 256, 512, 1024, 2048, 4096 }
To retain previous behavior, filter out 0 MiB (invalid size), 64 MiB
and 4096 MiB options.
Removing the floating point stuff also saves 1.5k from text segment:
clearfog : spl/u-boot-spl:all -1592 spl/u-boot-spl:text -1592
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dirk Eibach <dirk.eibach@gdsys.cc>
Cc: Stefan Roese <sr@denx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
During DDR-2133 operation, the transmit data eye margins determined
during the memory controller initialization may be sub-optimal, set
DEBUG_29[12] and DEBUG_29[13:16] = 4'b0100 before MEM_EN is set.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
The initial training for the DDRC may provide results that are not
optimized. The workaround provides better read timing margins.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Per the latest erratum document, update step 4 and step 8, only
DEBUG_29[21] is changed, all other bits should not be changed.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
The code uses a lot of signed numbers, which ended up in variables
of unsigned type, which resulted in all sorts of underflows. This
in turn caused incorrect calibration on certain boards. Moreover,
repair the readout of the DQ delay, which was being pulled from
wrong register.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
Just staticize global variables in sequencer, since there is no
point in having these symbols available outside of the DDR code.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
Originally, the DLEVEL selects the debug level within the sequencer code,
but only displays the messages on that particular debug level. Tweak the
handling such that for particular debug level, debug messages on that
level and lower are displayed. This allows better regulation of debug
message verbosity.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
This one last set of delay configuration registers was not properly
zeroed out originally, fix it and zero them out.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
There is no point in resetting the ODT setting if the write test
failed, since the code will always retry the calibration and thus
reconfigure the ODT anyway OR the code will fail calibration and
halt.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
Every invocation of the scc_mgr_set_dqs_en_delay_all_ranks() is
followed by SCC manager update. Moreover, only this function
triggers the SCC manager update internally. Thus, remove the
internal invocation to avoid triggering the update twice.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
The hi address bitfield in the protection rule must be set to
the last address in the region which the rule represents. The
behavior is now in-line with code generated by Quartus 15.1 .
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
The code should be setting registers to zero, not one register to value.
Swap the order of arguments to correct the behavior. The behavior is now
in-line with code generated by Quartus 15.1 .
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
In the most unlikely case the DQS tracking was to be disabled,
make sure we do not errornously re-enable it. Note that DQS
tracking is enabled on all systems observed thus far.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
The bit 22 is in fact DQS tracking enable bit (dqstrken) and there
is a macro for this bit already, so use it.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
Cc: Chin Liang See <clsee@altera.com>
Minor change: chosen is written with one "o".
No code change here, only comment & printf.
Signed-off-by: Alexander Merkle <alexander.merkle@lauterbach.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Armada 38x has a maximum of two cores. Probably copy/paste
bug from Armada XP.
Signed-off-by: Dirk Eibach <dirk.eibach@gdsys.cc>
Signed-off-by: Stefan Roese <sr@denx.de>
During initial DDR training, false parity errors may be detected.
This patch adds workaround to fix the erratum.
Tested on LS2085QDS and LS2080RDB.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Add support of address parity for DDR4 UDIMM or discrete memory.
It requires to configurate corresponding MR5[2:0] and
TIMING_CFG_7[PAR_LAT]. Parity can be turned on by hwconfig,
e.g. hwconfig=fsl_ddr:parity=on.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Correct spelling of "U-Boot" shall be used in all written text
(documentation, comments in source files etc.).
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Minkyu Kang <mk7.kang@samsung.com>
Following commit 61bd2f75, exclude unused DDR controller from
calculating RAM size for SPL boot.
Signed-off-by: Ed Swarthout <Ed.Swarthout@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Erratum A-009663 workaround requires to set DDR_INTERVAL[BSTOPRE] to 0
before setting DDR_SDRAM_CFG[MEM_EN] and set DDR_INTERVAL[BSTOPRE]
to the desired value after DDR initialization has completed.
When DDR controller is configured to operate in auto-precharge
mode(DDR_INTERVAL[BSTOPRE]=0), this workaround is not needed.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Reviewed-by: York Sun <york.sun@nxp.com>
During the receive data training, the DDRC may complete on a
non-optimal setting that could lead to data corruption or
initialization failure.
Workaround: before setting MEM_EN, set DEBUG_29 register with
specific value for different data rates.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Reviewed-by: York Sun <york.sun@nxp.com>
In a number of places we had wordings of the GPL (or LGPL in a few
cases) license text that were split in such a way that it wasn't caught
previously. Convert all of these to the correct SPDX-License-Identifier
tag.
Signed-off-by: Tom Rini <trini@konsulko.com>
This should make it clear that this symbol is meant to be defined by
board headers.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Acked-by: Stefan Roese <sr@denx.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
If MV_DEBUG_WL is defined, DEBUG_WL_S and DEBUG_WL_D macros are missing.
In addition to that, get rid of debug output printing non-existent
counter variable.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Acked-by: Stefan Roese <sr@denx.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
Currently, ECC support is enabled for all Armada XP boards. So the
DDR3 driver tries to configure the controller with ECC support, even
on boards without ECC. This patch makes this ECC optional which now
can be configured on a board-per-board basis.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Luka Perkov <luka.perkov@sartura.hr>
Cc: Phil Sutter <phil@nwl.cc>
gcc 5.1 generates this new warning (for Armada 38x platforms):
drivers/ddr/marvell/a38x/ddr3_debug.c: In function 'hws_ddr3_tip_read_training_result':
drivers/ddr/marvell/a38x/ddr3_debug.c:177:40: warning: 'sizeof' on array
function parameter 'result' will return size of 'enum hws_result (*)[1]' [-Wsizeof-array-argument]
memcpy(result, training_result, sizeof(result));
^
drivers/ddr/marvell/a38x/ddr3_debug.c:171:31: note: declared here
u32 dev_num, enum hws_result result[MAX_STAGE_LIMIT][MAX_INTERFACE_NUM])
^
Since this functions is not referenced anywhere, lets just remove it.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Luka Perkov <luka.perkov@sartura.hr>
As the errata A008336 and A008514 do not apply to all LS series SoCs
(such as LS1021A, LS1043A) we move them to an soc specific file
Signed-off-by: Yuan Yao <yao.yuan@freescale.com>
Reviewed-by: York Sun <yorksun@freescale.com>
DDR errata-A008378 applies to LS1021-20-22A-R1.0, T1023-R1.0,
T1024-R1.0, T1040-42-20-22-R1.0/R1.1, it has been fixed on
LS102x Rev2.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Reviewed-by: York Sun <yorksun@freescale.com>
In case four chip-selects are all active, the turnaround times need to
increase to avoid overlapping under heavy load.
Signed-off-by: York Sun <yorksun@freescale.com>
The workaround requires different setting for range 1 vs 2.
Also adjust timeout value for waiting for controller to be idle.
Signed-off-by: York Sun <yorksun@freescale.com>
BIST test code has a typo, resulting the binding registers not
maintained as expected. This typo results BIST runs twice on
the covered memory.
Signed-off-by: York Sun <yorksun@freescale.com>
Reported-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Freescale LSCH3 platforms use two DDR controlers interleaving mode out of
reset. It can be configured to disable one controller. To support this
operation, the driver needs to detect and skip the disabled controller.
Signed-off-by: York Sun <yorksun@freescale.com>
Freescale's LS2085A is a another personality of LS2080A SoC with
support of AIOP and DP-DDR.
This Patch adds support of LS2085A Personality.
Signed-off-by: Pratiyush Mohan Srivastava <pratiyush.srivastava@freescale.com>
Signed-off-by: Prabhakar Kushwaha <prabhakar@freescale.com>
[York Sun: Updated MAINTAINERS files
Dropped #ifdef in cpu.h
Add CONFIG_SYS_NS16550=y in defconfig]
Reviewed-by: York Sun <yorksun@freescale.com>
LS2080A is a prime personality of Freescale’s LS2085A. It is a non-AIOP
personality without support of DP-DDR, L2 switch, 1588, PCIe endpoint etc.
So renaming existing LS2085A code base to reflect LS2080A (Prime personality)
Signed-off-by: Pratiyush Mohan Srivastava <pratiyush.srivastava@freescale.com>
Signed-off-by: Prabhakar Kushwaha <prabhakar@freescale.com>
[York Sun: Dropped #ifdef in cpu.c for cpu_type_list]
Reviewed-by: York Sun <yorksun@freescale.com>
SAR1_CPU_CORE_MASK was wrong, probably copy/paste
from another architecture.
Signed-off-by: Dirk Eibach <dirk.eibach@gdsys.cc>
Reviewed-by: Stefan Roese <sr@denx.de>
A lot of extra configuration information was left over in the
Marvell serdes and DDR3 initialization code for boards that
U-boot does not support. Remove this extra config information,
and the concept of fixing up board topologies with information
loaded from an EEPROM. If this needs to be done, it should be
handled in the board file, not in core code.
Signed-off-by: Kevin Smith <kevin.smith@elecsyscorp.com>
Acked-by: Stefan Roese <sr@denx.de>
Cc: Dirk Eibach <eibach@gdsys.de>
Cc: Luka Perkov <luka.perkov@sartura.hr>
After consulting with some of the SPDX team, the conclusion is that
Makefiles are worth adding SPDX-License-Identifier tags too, and most of
ours have one. This adds tags to ones that lack them and converts a few
that had full (or in one case, very partial) license blobs into the
equivalent tag.
Cc: Kate Stewart <kstewart@linuxfoundation.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
SR_IE(Self-refresh interrupt enable) is needed for
Hardware Based Self-Refresh. Make it configurable and let
board code handle the rest.
Signed-off-by: Joakim Tjernlund <joakim.tjernlund@transmode.se>
Reviewed-by: York Sun <yorksun@freescale.com>
introduce BIT() definition, used in at91_udc gadget
driver.
Signed-off-by: Heiko Schocher <hs@denx.de>
[remove all other occurrences of BIT(x) definition]
Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
Acked-by: Stefan Roese <sr@denx.de>
Acked-by: Anatolij Gustschin <agust@denx.de>
Fix the following problem:
drivers/ddr/altera/sequencer.c: In function 'sdram_calibration_full':
drivers/ddr/altera/sequencer.c:1943:25: warning: 'found_failing_read' may be used uninitialized in this function [-Wmaybe-uninitialized]
if (found_passing_read && found_failing_read)
^
drivers/ddr/altera/sequencer.c:1803:26: note: 'found_failing_read' was declared here
u32 found_passing_read, found_failing_read, initial_failing_dtap;
^
Signed-off-by: Marek Vasut <marex@denx.de>
This gem is really really rare, there was an actual float used in
the Altera DDR init code, which pulled in floating point ops from
the libgcc, just wow.
Since we don't support floating point operations the same way Linux
does not support them, replace this with an integer multiplication
and division combo. This removes some 2kiB of size from the SPL as
the floating point ops are no longer pulled in from libgcc.
This was detected by enabling CONFIG_USE_PRIVATE_LIBGCC=y , which
does not contain the floating point bits.
Signed-off-by: Marek Vasut <marex@denx.de>
This patch introduces the SDRAM scrubbing for ECC enabled board
to fill/initialize the ECC bytes. This is done via the XOR engine
to speed up the process. The scrubbing is a 2-stage process:
1) SPL scrubs the area 0 - 0x100.0000 (16MiB) for the main U-Boot
2) U-Boot scrubs the remaining SDRAM area(s)
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Luka Perkov <luka.perkov@sartura.hr>
This is tested on the DB-MV784MP-GP eval board. To really enable ECC
support on this board the I2C EEPROM needs to get changed. As it
saves the enabling of ECC support internally. For this the following
commands can be used to enable ECC support on this board:
Its recommended for first save (print) the value(s) in this EEPROM
address:
=> i2c md 4e 0.1 2
0000: 05 00 ..
To enable ECC support you need to set bit 1 in the 2nd byte:
Marvell>> i2c mw 4e 1.1 02
Marvell>> i2c md 4e 0.1 2
0000: 05 02 ..
To disable ECC support again, please use this command:
Marvell>> i2c mw 4e 1.1 00
Marvell>> i2c md 4e 0.1 2
0000: 05 00 ..
On other AXP boards, simply plugging an ECC DIMM should be enough to
enable ECC support.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Luka Perkov <luka.perkov@sartura.hr>
Fix most of the dangling checkpatch issues, no functional change.
There are still 7 warnings, 1 checks , but those are left in place
for the sake of readability of the code.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Actually convert the sequencer code to use socfpga_sdram_misc_config
instead of the various macros. This is just an sed exercise here, no
manual coding needed.
This patch actually removes the need to include any board-specific
files in sequencer.c , so sequencer.c namespace is now no longer
poluted by QTS-generated macros.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
This is another macro used to obfuscate the real code. The
T(INIT|RESET)_CNTR._VAL is always defined, so this indirection
is unnecessary. Get rid of this.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Just use READ_VALID_FIFO_SIZE directly, no need for this macro obfuscation.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Introduce structure socfpga_sdram_misc_config to wrap the remaining
misc configuration values in board file. Again, introduce a function,
socfpga_get_sdram_misc_config(), which returns this the structure. This
is almost the final step toward wrapping the nasty QTS generated macros
in board files and reducing the pollution of the namespace.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Actually convert the sequencer code to use socfpga_sdram_io_config
instead of the IO_* macros. This is just an sed excercise here, no
manual coding needed.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Introduce structure socfpga_sdram_io_config to wrap the IO configuration
values in board file. Introduce socfpga_get_sdram_io_config() function,
which returns this the structure. This is another step toward wrapping
the nasty QTS generated macros in board files and reducing the pollution
of the namespace.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Actually convert the sequencer code to use socfpga_sdram_rw_mgr_config
instead of the RW_MGR_* macros. This is just an sed exercise here, no
manual coding needed.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Introduce structure socfpga_sdram_rw_mgr_config to wrap the RW manager
configuration values in board file. Introduce a complementary function,
socfpga_get_sdram_rwmgr_config(), which returns this the structure.
This is another step toward wrapping the nasty QTS generated macros
in board files and reducing the pollution of the namespace.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Introduce two wrapper functions, socfpga_get_seq_ac_init() and
socfpga_get_seq_inst_init() to avoid direct inclusion of the
sequencer_auto_ac_init.h and sequencer_auto_inst_init.h QTS
generated files. This reduces namespace pollution again.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
This is defined in the QTS-generated headers, so it must not be
re-defined in sequencer.h .
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
These parameters are not used in the code, zap them and the
macros which are used by them as well.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Move the files generated by QTS into the board directory, they should not
be part of the driver files at all.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Originally, both sdram_start and sdram_end were 64b values. The
sdram_start had no reason for being so, since our address space
is only 32b, so switching sdram_start to u32 is simple.
The sdram_end is a bit more complex, since it can actually be
set to (1 << 32) if someone really wanted to use an SoCFPGA with
4 GiB of DRAM and fixed the code around a little. But, the code
handling the protection rules internally decrements the sdram_end
variable anyway. Thus, instead of calling the code and passing in
the address of the SDRAM end, pass in the address already decremented
by one. This lets the sdram_end be 32b as well.
Signed-off-by: Marek Vasut <marex@denx.de>
Clean the function up so that it's obvious what it is doing,
fix the formating strings in debug outputs, add kerneldoc.
Make the function return proper errno-compliant return values
and propagate this change throughout sdram.c
Signed-off-by: Marek Vasut <marex@denx.de>
Clean up coding style, mostly clean up comments, add kerneldoc.
Also, zap assignment of the "cs" variable, which is outright
dead code, so just remove it.
Signed-off-by: Marek Vasut <marex@denx.de>
Pluck out all of the CONFIG_HPS_SDR_CTRLCFG_* macros. This change
makes sdram.c completely clear of these macros and allows removing
of the ugly include of sdram.h . The namespace is now a much nicer
place!
Signed-off-by: Marek Vasut <marex@denx.de>
Introduce socfpga_sdram_get_config() function implement in a board file,
which returns the socfpga_sdram_config structure. This is the last step
in cleaning up the socfpga_mmr_init_full(), but not the last step which
allows removing the inclusion of sdram.h from drivers/ddr/altera/sdram.c
thus far.
Signed-off-by: Marek Vasut <marex@denx.de>
Rework remaining two register setting functions such that they only
return the final register value. Move the register setting into the
block of register I/O in sdram_mmr_init_full().
Signed-off-by: Marek Vasut <marex@denx.de>
Merge sdr_set_*() functions which are just setting registers among
the sea of register setting in sdram_mmr_init_full(). There is no
need to keep them separate this way, there is nothing special about
them.
Signed-off-by: Marek Vasut <marex@denx.de>
Suck out all the CONFIG_HPS_SDR_CTRLCFG_* from sdram_mmr_init_full()
into the socfpga_sdram_config structure. There is still one ugly
macro left behind, but this will be taken care of in subsequent patch.
Signed-off-by: Marek Vasut <marex@denx.de>
Zap all the ad-hoc readbacks from the registers and other useless
and broken debug output. This is really not useful and is only
confusing.
Signed-off-by: Marek Vasut <marex@denx.de>
Introduce this seemingly massive structure, which holds required values
of all the registers of the SDRAM controller. The idea here is to avoid
including the sdram.h header file, which is full of ad-hoc macros that
polute the global namespace. Once the cleanup of sdram.c would be complete
and all registers would be loaded from this new socfpga_sdram_config, a
board file will only pass this structure into the sdram.c . This will
hide all the horrors generated by QTS in the board directory.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
Merge set_sdr_dram_timing{1,2,3,4,lowpwr}() into single function
set_sdr_dram_timing() , since there's no point in keeping all this
stuff separate anymore.
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of the constant clrsetbits_le32(), instead prepare the whole
content of the register once and write it at the end of the function.
The big plan here is to remove all the CONFIG_HPS_SDR_ macros, hide
them in QTS compatibility layer in board implementation and pass only
a small structure into the driver.
Signed-off-by: Marek Vasut <marex@denx.de>
Remove all parameters of this function, since they are only constants
passed in from another function, so make them local. Also, rename the
function to get_errata_rows() as this is closer to what it does.
Signed-off-by: Marek Vasut <marex@denx.de>
This function is implementing the DDR calibration Stage 3 as
described in Altera EMI_RM 2015.05.04 . The main body of this
function is almost identical to Stage 1.3 (DQ/DQS centering)
for all but two flags -- use_read_test and update_fom. Convert
this function to call rw_mgr_mem_calibrate_dq_dqs_centering()
with the correct flags set to trim down the code duplication.
Moreover, reorder the remnants in the function a little and
convert the function to return either 0 or -EIO in case of
success and failure respectively, to match the common return
value convention.
Signed-off-by: Marek Vasut <marex@denx.de>
Clean up coding style, mostly broken multiline formatting strings
in debug output. Also clean up data types and constify where applicable.
Signed-off-by: Marek Vasut <marex@denx.de>
Fix data types and constify where applicable, fix broken multiline
debug strings and fix comments. No functional change.
Signed-off-by: Marek Vasut <marex@denx.de>
Fix the return value of the function to match common convention
where 0 means success and negative means error. Fix the return
values in case of an error to use errno.h codes.
Signed-off-by: Marek Vasut <marex@denx.de>
The read_group and write_group arguments are the same in all
cases when this function is invoked, just merge them into one
rw_group argument. Also, clean up the function argument data
types and constify them.
Signed-off-by: Marek Vasut <marex@denx.de>
The search_stop_check() always resets the bit_chk to a new value,
so bit_chk will never survive between search_left_edge() and
search_right_edge() invocation. There is thus no need to pass
it into these functions, so make bit_chk local to them.
Signed-off-by: Marek Vasut <marex@denx.de>
Factor out center_dq_windows(), which is common code between
stage 2 and stage 3 of the calibration again and cater for
the minor differences.
Signed-off-by: Marek Vasut <marex@denx.de>
Factor out the get_window_mid_index() for finding the best DQ
window centering. This is again code common to two stages, so
pull it out.
Signed-off-by: Marek Vasut <marex@denx.de>
First of all, don't break strings in the debug_cond() calls across
multiple lines. Next, zap a couple of really odd casts in the code.
Finally, make i into a signed variable, so it is possible to avoid
constructs of the form:
for (i = FOO;; i--) { <code>; if (i == 0) break; }.
Signed-off-by: Marek Vasut <marex@denx.de>
Factor out common code from search_left_edge() and search_right_edge()
which checks whether searching for the window edge should stop. The
code is almost identical, so pull it into separate function and cater
for the minor differences.
Signed-off-by: Marek Vasut <marex@denx.de>
Factor out common code from rw_mgr_mem_calibrate_writes_center()
and rw_mgr_mem_calibrate_vfifo_center() for searching for the
left edge of the window. The code is almost identical, so pull
it into separate function and cater for the minor differences.
Signed-off-by: Marek Vasut <marex@denx.de>
Factor out common code from rw_mgr_mem_calibrate_writes_center()
and rw_mgr_mem_calibrate_vfifo_center() for searching for the
right edge of the window. The code is almost identical, so pull
it into separate function and cater for the minor differences.
Signed-off-by: Marek Vasut <marex@denx.de>
Reorder the content of the function a little and fix the comments
so they at least become full sentences. Constify function args.
Fix the return value to match the common convention of 0 meaning
success.
Signed-off-by: Marek Vasut <marex@denx.de>
The C language has signed types, so make use of them. Fix this
obscene loop, which tries to do away with unsigned type, but
just makes the code more cryptic.
Signed-off-by: Marek Vasut <marex@denx.de>
Clean up the debug output handling at the end of the function
and factor out common function call from the condition.
Signed-off-by: Marek Vasut <marex@denx.de>
Implement common sdr_find_phase_delay() function and convert all
places which use such code to this unified function. No functional
change to the code.
Signed-off-by: Marek Vasut <marex@denx.de>
If d != 0 after sdr_working_phase() finishes, the else branch
of the condition has no impact on anything at all, since work_end
was already set independently of the value of "d" . Zap this
useless code.
Signed-off-by: Marek Vasut <marex@denx.de>
Fix data types, constify where applicable, fix comments and debug
strings. In case enough failing reads are detected, just return
right away instead of breaking from the loop. This lets us make
the error handling at the end much simpler.
Signed-off-by: Marek Vasut <marex@denx.de>
Pluck out all this VFIFO value counting, which turns out to be
completely unused. Thus, remove it to simplify the code.
Signed-off-by: Marek Vasut <marex@denx.de>
The "d" variable is set to 0 in sdr_backup_phase() and is not used
at all in sdr_nonworking_phase(). Make it local and zap it respectively.
Signed-off-by: Marek Vasut <marex@denx.de>
Wrap dtaps_per_ptap into the sdr_working_phase() function to trim
down the number of params. It's a constant too.
Signed-off-by: Marek Vasut <marex@denx.de>
It is now clear that the max_working_cnt variable is totally unused.
Zap the variable and zap other variables which became unused due to
this change too.
Signed-off-by: Marek Vasut <marex@denx.de>
Pull out the loop for eaching working/non-working DQS enable phase
into a separate function, as this is mostly common code between.
Clean up sdr_working_phase() and sdr_nonworking_phase() while switching
these two functions to the common sdr_find_phase().
Signed-off-by: Marek Vasut <marex@denx.de>
Get rid of found_{begin,end} variables. Instead of breaking out
through all of the loops, just return when the begin/end of the
window is found and be done with it. Also clean up the trailing
conditional expression, which is now much easier.
Signed-off-by: Marek Vasut <marex@denx.de>
Fix the arguments passed to these functions. The bit_chk is
overriden by rw_mgr_mem_calibrate_read_test_all_ranks() which
is invoked by all three sdr_*_phase() functions, so just make
this into local variable.
Signed-off-by: Marek Vasut <marex@denx.de>
Fix the arguments passed to these functions. The grp argument
does not have to be passed via reference, it's never modified
within either of those functions, so make it into a value.
Signed-off-by: Marek Vasut <marex@denx.de>
Reorder the end of the function a little by moving the conditional
debug output around a little. Rename the function from _centre() to
_center(). Document the function in kerneldoc.
Signed-off-by: Marek Vasut <marex@denx.de>
This function is a treasure trove of ad-hoc iterative
implementations of mathematical functions. Replace all
of those with their non-iterative counterpart.
Signed-off-by: Marek Vasut <marex@denx.de>
Clean up the arguments of this function. Most of the pointers
passed into the function are either not needed at all, or can
be passed as value instead of reference. Also fix the broken
multiline debug strings. No functional change.
Signed-off-by: Marek Vasut <marex@denx.de>
This function is only invoked from rw_mgr_mem_calibrate_dqs_enable_calibration()
and at this point, it is just one level of indirection, so wrap the
rw_mgr_mem_calibrate_vfifo_find_dqs_en_phase_sweep_dq_in_delay() into
rw_mgr_mem_calibrate_dqs_enable_calibration() to get rid of the level
of indirection.
Signed-off-by: Marek Vasut <marex@denx.de>
Replace at least one of the loops in this function with call of a
standard function call instead of the ad-hoc implementation. The
other one cannot be replaced, since the delay is incremented for
each group.
Signed-off-by: Marek Vasut <marex@denx.de>
The read_group and write_group params have the same value for all (one)
invocations of this function, just merge them into a single param.
Signed-off-by: Marek Vasut <marex@denx.de>
Start cleaning up this function. In the first part, just fix
the incorrectly broken debug strings and fix return value to
respect the common convention.
Signed-off-by: Marek Vasut <marex@denx.de>
Rework this function such that the code is more readable. Zap
unused parameter "num_tries" while at it. Also wrap parameter
"bit_chk" into this function as it's value is not used outside.
Finally, fix the return value from this function to match the
common expectation, where 0 means success.
Signed-off-by: Marek Vasut <marex@denx.de>
This function is called from one single place and it's sole purpose
is to call one single function with slightly modified arguments.
Zap this function to skip this useless intermediate step.
Signed-off-by: Marek Vasut <marex@denx.de>
Just extract this piece of functionality into separate function
to make the code better separated. This matches the division in
Altera documentation, Altera EMI_RM 2015.05.04 , section 1, the
UniPHY Calibration Stages.
Signed-off-by: Marek Vasut <marex@denx.de>
