The branch_if_master macro jumps to a label if the CPU is the "master"
core, which we define as having all affinity levels set to 0. To check
for this condition, we need to mask off some bits from the MPIDR
register, then compare the remaining register value against zero.
The implementation of this was slighly broken (it preserved the upper
RES0 bits), overly complicated and hard to understand, especially since
it lacked comments. The same was true for the very similar
branch_if_slave macro.
Use a much shorter assembly sequence for those checks, use the same
masking for both macros (just negate the final branch), and put some
comments on them, to make it clear what the code does.
This allows to drop the second temporary register for branch_if_master,
so we adjust all call sites as well.
Also use the opportunity to remove a misleading comment: the macro
works fine on SoCs with multiple clusters. Judging by the commit
message, the original problem with the Juno SoC stems from the fact that
the master CPU *can* be configured to be from cluster 1, so the
assumption that the master CPU has all affinity values set to 0 does not
hold there. But this is already mentioned above in a comment, so remove
the extra comment.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The ARMv8 architecture describes the "SError interrupt" as the fourth
kind of exception, next to synchronous exceptions, IRQs, and FIQs.
Those SErrors signal exceptional conditions from which the system might
not easily recover, and are normally generated by the interconnect as a
response to some bus error. A typical situation is access to a
non-existing memory address or device, but it might be deliberately
triggered by a device as well.
The SError interrupt replaces the Armv7 asynchronous abort.
Trusted Firmware enters U-Boot (BL33) typically with SErrors masked,
and we never enable them. However any SError condition still triggers
the SError interrupt, and this condition stays pending, it just won't be
handled. If now later on the Linux kernel unmasks the "A" bit in PState,
it will immediately take the exception, leading to a kernel crash.
This leaves many people scratching their head about the reason for
this, and leads to long debug sessions, possibly looking at the wrong
places (the kernel, but not U-Boot).
To avoid the situation, just unmask SErrors early in the ARMv8 boot
process, so that the U-Boot exception handlers reports them in a timely
manner. As SErrors are typically asynchronous, the register dump does
not need to point at the actual culprit, but it should happen very
shortly after the condition.
For those exceptions to be taken, we also need to route them to EL2,
if U-Boot is running in this exception level.
This removes the respective code snippet from the Freescale lowlevel
routine, as this is now handled in generic ARMv8 code.
Reported-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The BL31 expects the GIC to be uninitialized. Thus, if we are loading
the BL31 by the SPL we must not initialize it. If u-boot is loaded by
the SPL directly, it will initialize the GIC again (in the same
lowlevel_init()).
This was tested on a custom board with SPL loading the BL31 and jumping
to u-boot as BL33 as well as loading u-boot directly by the SPL. In case
the ATF BL1/BL2 is used, this patch won't change anything, because no
SPL is used at all.
Signed-off-by: Michael Walle <michael@walle.cc>
Move it out of lowlevel.S into spintable.S. On layerscape, the secondary
CPUs are brought up in main u-boot. This will make it possible to only
compile the spin table code for the main u-boot and omit it in SPL.
This saves about 720 bytes in the SPL.
Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Priyanka Jain <priyanka.jain@nxp.com>
The generic armv8 code already has support to bring up the secondary
cores. Thus, don't hardcode the jump in the layerscape lowlevel_init to
the spin table code; instead just return early and let the common armv8
code handle the jump. This way we can actually use the CPU_RELEASE_ADDR
feature.
Signed-off-by: Michael Walle <michael@walle.cc>
[Rebased, Removed kontron_sl28.h change as file does not exist]
Signed-off-by: Priyanka Jain <priyanka.jain@nxp.com>
Spin tables are broken with bootefi. This is because - in contrast to
the booti call chain - there is no call to smp_kick_all_cpus(). Due to
this missing call the secondary CPUs are never released from their "wait
for interrupt state", see secondary_boot_func() in lowlevel.S.
Originally, this "wait for interrupt" is there to make sure, the spin
table is cleared before the secondary cores read it for the first time.
But the boot flow for the layerscape architecture is different from
that. The CPUs are release from their BootROM _after_ U-Boot's
spin-table is cleared, see fsl_layerscape_wake_seconday_cores() in mp.c.
Thus, there is no need to wait for this interrupt and no need for
kicking all cores on cpu_release. An atomic 64bit write to the
spin-table and a "sev" is sufficient.
Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Priyanka Jain <priyanka.jain@nxp.com>
Flushing L3 cache may need variable time depending upon cache line
allocation.
Coming up with a proper timeout value would be best handled by
simulations under multiple scenarios in your actual system.
>From the purely HN-F point of view, the flush would take ~15 cycles for
a clean line, and ~22 cycles for a dirty line. For the dirty line case,
there are many variables outside the HN-F that will increase the
duration per line. For example, a *DBIDResp from the SN-F/SBSX,
memory controller latency, SN-F/SBSX RetryAck responses, CCN ring
congestion, CCN ring hops, etc, etc. The worst-case timeout would
have to factor in all of these variables plus the HN-F cycles for
every line in the L3, and assuming all lines are dirty
In case if L3 is not flushed properly, system behaviour will be
erratic, so remove timeout and add loop to check status of L3 cache.
System will stuck in while loop if there is some issue in L3 cache
flushing.
Signed-off-by: Udit Kumar <udit.kumar@nxp.com>
Signed-off-by: Meenakshi Aggarwal <meenakshi.aggarwal@nxp.com>
Reviewed-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
OCRAM initialization is performed by TFA, Hence
skipped from u-boot.
Signed-off-by: Ruchika Gupta <ruchika.gupta@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
In case SError happens at EL2, if SCR_EL3[EA] is not routing it to
EL3, and SCR_EL3[RW] is set to aarch64, setting HCR_EL2[AMO] routes
the exception to EL2. Otherwise this exception is not taken.
Signed-off-by: York Sun <york.sun@nxp.com>
TZASC controller configurations are similar. Put them in a macro and
avoid code duplication.
Signed-off-by: Priyanka Jain <priyanka.jain@nxp.com>
Signed-off-by: Sriram Dash <sriram.dash@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
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>
Check LS1043A/LS2080a by device ID without using personality ID to
determine revision number. This check applies to all various
personalities of the same SoC family.
Signed-off-by: Wenbin Song <wenbin.song@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
CCN-504 HPF registers were believed to be accessible only from EL3.
However, recent tests proved otherwise. Remove checking for exception
level to re-enable L3 cache flushing for all levels.
Signed-off-by: York Sun <york.sun@nxp.com>
Tested-by: Zhao Qiang <qiang.zhao@nxp.com>
LS2080 family has CCN-504 cache coherent interconnet. Other SoCs
in LSCH3 family may have differnt interconnect.
Signed-off-by: Ashish Kumar <Ashish.Kumar@nxp.com>
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
[YS: revised commit message]
Reviewed-by: York Sun <york.sun@nxp.com>
The duplicate definitions for IH_ARCH_ARM and IH_ARCH_ARM64 are removed.
The definitions in <image.h> are used.
According to this modification, the comparison between os arch and cpu
arch is done in C programming instead of ASM programming.
Signed-off-by: Alison Wang <alison.wang@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
When U-Boot boots from EL2, skip some lowlevel init code requiring
EL3, including CCI-400/CCN-504, trust zone, GIC, etc. These
initialization tasks are carried out before U-Boot runs. This applies
to the RAM version image used for SPL boot if PPA is loaded first.
Signed-off-by: York Sun <york.sun@nxp.com>
Number of TZASC instances may vary across NXP SoCs.
So put TZASC configuration under instance specific defines.
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Signed-off-by: Ashish Kumar <Ashish.Kumar@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
For 64-bit kernel, there is a warning about x1-x3 nonzero in violation
of boot protocol. To fix this issue, input argument 4 is added for
armv8_switch_to_el2 and armv8_switch_to_el1. The input argument 4 will
be set to the right value, such as zero.
Signed-off-by: Alison Wang <alison.wang@nxp.com>
Reviewed-by: Alexander Graf <agraf@suse.de>
Tested-by: Ryan Harkin <ryan.harkin@linaro.org>
Tested-by: Michal Simek <michal.simek@xilinx.com>
Reviewed-by: York Sun <york.sun@nxp.com>
The LS1043A rev1.1 silicon supports two types of GIC offset: 4K
alignment and 64K alignment. The bit SCFG_GIC400_ALIGN[GIC_ADDR_BIT]
is used to choose which offset will be used.
The LS1043A rev1.0 silicon only supports the CIG offset with 4K
alignment.
If GIC_ADDR_BIT bit is set, 4K alignment is used, or else 64K alignment
is used. 64K alignment is the default setting.
Overriding the weak smp_kick_all_cpus, the new impletment is able to
detect GIC offset.
The default GIC offset in kernel device tree is using 4K alignment, it
need to be fixed if 64K alignment is detected.
Signed-off-by: Wenbin Song <wenbin.song@nxp.com>
Signed-off-by: Mingkai Hu <mingkai.hu@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
Clear the content to zero and the ECC error bit of OCRAM1/2.
The OCRAM must be initialized to ZERO by the unit of 8-Byte before
accessing it, or else it will generate ECC error. And the IBR has
accessed the OCRAM before this initialization, so the ECC error
status bit should to be cleared.
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Pratiyush Srivastava <pratiyush.srivastava@nxp.com>
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
It is recommended to set forced-order mode in RNI-6,
RNI-20 for performance optimization in LS2088A.
Both LS2080A, LS2088A families has CONFIG_LS2080A define.
As above update is required only for LS2088A, skip this
for LS2080A SoC family.
Signed-off-by: Priyanka Jain <priyanka.jain@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
To support loading a 32-bit OS, the execution state will change from
AArch64 to AArch32 when jumping to kernel.
The architecture information will be got through checking FIT image,
then U-Boot will load 32-bit OS or 64-bit OS automatically.
Signed-off-by: Ebony Zhu <ebony.zhu@nxp.com>
Signed-off-by: Alison Wang <alison.wang@nxp.com>
Signed-off-by: Chenhui Zhao <chenhui.zhao@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
LS2080 SoC and its personalities does not support TZASC
But other new SoCs like LS2088A, LS1088A supports TZASC
Hence, skip initializing TZASC for Ls2080A based on SVR
Signed-off-by: Priyanka Jain <priyanka.jain@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
TZASC registers like TZASC_GATE_KEEPER, TZASC_REGION_ATTRIBUTES
are 32-bit regsiters.
So while doing register load-store operations, 32-bit intermediate
register, w0 should be used.
Update x0 register to w0 register type.
Signed-off-by: Priyanka Jain <priyanka.jain@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
SoC-specific logic may be required for all forms of cache-wide
operations; invalidate and flush of both dcache and icache (note that
only 3 of the 4 possible combinations make sense, since the icache never
contains dirty lines). This patch adds an optional hook for all
implemented cache-wide operations, and renames the one existing hook to
better represent exactly which operation it is implementing. A dummy
no-op implementation of each hook is provided.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Move this option to Kconfig and clean up existing uses.
Signed-off-by: York Sun <york.sun@nxp.com>
CC: Mingkai Hu <mingkai.hu@nxp.com>
CC: Gong Qianyu <Qianyu.Gong@nxp.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
According to design specification, the L2 cache operates at the same
frequency as the A72 CPUs in the cluster with a 3-cycle latency, so
increase the L2 Data RAM and Tag RAM latency to 3 cycles, or else,
will run into different call trace issues.
Signed-off-by: Mingkai Hu <mingkai.hu@nxp.com>
Signed-off-by: Gong Qianyu <Qianyu.Gong@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
The QorIQ LS1012A processor, optimized for battery-backed or
USB-powered, integrates a single ARM Cortex-A53 core with a hardware
packet forwarding engine and high-speed interfaces to deliver
line-rate networking performance.
This patch add support of LS1012A SoC along with
- Update platform & DDR clock read logic as per SVR
- Define MMDC controller register set.
- Update LUT base address for PCIe
- Avoid L3 platform cache compilation
- Update USB address, errata
- SerDes table
- Added CSU IDs for SDHC2, SAI-1 to SAI-4
Signed-off-by: Calvin Johnson <calvin.johnson@nxp.com>
Signed-off-by: Makarand Pawagi <makarand.pawagi@mindspeed.com>
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
It is not mandatory for Layerscape SoCs to have SMMU. SoCs like
LS1012A are layerscape SoC without SMMU IP.
So put SMMU configuration code under SMMU_BASE.
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
LS2080A is the primary SoC, and LS2085A is a personality with AIOP
and DPAA DDR. The RDB and QDS boards support both personality. By
detecting the SVR at runtime, a single image per board can support
both SoCs. It gives users flexibility to swtich SoC without the need
to reprogram the board.
Signed-off-by: York Sun <york.sun@nxp.com>
CC: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Reviewed-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Enable wuo config to accelerate coherent ordered writes for LS2080A
and LS2085A.
WRIOP IP is connected to RNI-20 Node.
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>
There are two LS series processors are built on ARMv8 Layersacpe
architecture currently, LS2085A and LS1043A. They are based on
ARMv8 core although use different chassis, so create fsl-layerscape
to refactor the common code for the LS series processors which also
paves the way for adding LS1043A platform.
Signed-off-by: Mingkai Hu <Mingkai.Hu@freescale.com>
Signed-off-by: Hou Zhiqiang <B48286@freescale.com>
Signed-off-by: Gong Qianyu <Qianyu.Gong@freescale.com>
Reviewed-by: York Sun <yorksun@freescale.com>
2015-10-29 10:34:00 -07:00
Renamed from arch/arm/cpu/armv8/fsl-lsch3/lowlevel.S (Browse further)
