Commit 643aae1406
deleted include/linux/config.h but missed to
delete _LINUX_CONFIG_H macro.
It is no longer used at all.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Erratum A-005812 Incorrect reservation clearing in Write Shadow mode can
result in invalid atomic operations. For u-boot, this erratum only impacts
SoCs running in write shadow mode.
Signed-off-by: York Sun <yorksun@freescale.com>
To align with chassis generation 2 spec, all cores are numbered in sequence.
The cores may reside across multiple clusters. Each cluster has zero to four
cores. The first available core is numbered as core 0. The second available
core is numbered as core 1 and so on.
Core clocks are generated by each clusters. To identify the cluster of each
core, topology registers are examined.
Cluster clock registers are reorganized to be easily indexed.
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
The PIR parsing algorithm we used is not only for E6500. It applies to all
SoCs with chassis 2.
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Makes it a bit easier to see if we've properly set them. While
we're in there, modify the accesses to HDBCR0 and HDBCR1 to actually
use those definitions.
Signed-off-by: Andy Fleming <afleming@freescale.com>
By extracting these defines into a header, they can be re-used by other
C sources as well. This will be done by the SPL framework OS boot
support.
Signed-off-by: Stefan Roese <sr@denx.de>
Once u-boot sets the spin table to cache-enabled memory, old kernel which
uses cache-inhibit mapping without coherence will not work properly. We
use this temporary fix until kernel has updated its spin table code.
For now this fix is activated by default. To disable this fix for new
kernel, set environmental variable "spin_table_compat=no". After kernel
has updated spin table code, this default shall be changed.
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Move spin table to cached memory to comply with ePAPR v1.1.
Load R3 with 64-bit value if CONFIG_SYS_PPC64 is defined.
'M' bit is set for DDR TLB to maintain cache coherence.
See details in doc/README.mpc85xx-spin-table.
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
R6 was in ePAPR draft version but was dropped in official spec.
Removing it to comply.
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Using E6500 L1 cache as initram requires L2 cache enabled.
Add l2-cache cluster enabling.
Setup stash id for L1 cache as (coreID) * 2 + 32 + 0
Setup stash id for L2 cache as (cluster) * 2 + 32 + 1
Stash id for L2 is only set for Chassis 2.
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Commit 709389b6 unintentionally used the Unicode version of the
apostrophy. Replace it with the normal ASCII version.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
For the cores with multiple threads, we need to figure out which physical
core a thread belongs. To match the core ids, update PIR registers and
spin tables.
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Erratum A004510 says that under certain load conditions, modified
cache lines can be discarded, causing data corruption.
To work around this, several CCSR and DCSR register updates need to be
made in a careful manner, so that there is no other transaction in
corenet when the update is made.
The update is made from a locked cacheline, with a delay before to flush
any previous activity, and a delay after to flush the CCSR/DCSR update.
We can't use a readback because that would be another corenet
transaction, which is not allowed.
We lock the subsequent cacheline to prevent it from being fetched while
we're executing the previous cacheline. It is filled with nops so that a
branch doesn't cause us to fetch another cacheline.
Ordinarily we are running in a cache-inhibited mapping at this point, so
we temporarily change that. We make it guarded so that we should never
see a speculative load, and we never do an explicit load. Thus, only the
I-cache should ever fill from this mapping, and we flush/unlock it
afterward. Thus we should avoid problems from any potential cache
aliasing between inhibited and non-inhibited mappings.
NOTE that if PAMU is used with this patch, it will need to use a
dedicated LAW as described in the erratum. This is the responsibility
of the OS that sets up PAMU.
Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
This erratum applies to the following SoCs:
P4080 rev 1.0, 2.0, fixed in rev 3.0
P2041 rev 1.0, 1.1, fixed in rev 2.0
P3041 rev 1.0, 1.1, fixed in rev 2.0.
Workaround for erratum NMG_CPU_A011 is enabled by default. This workaround
may degrade performance. P4080 erratum CPU22 shares the same workaround.
So it is always enabled for P4080. For other SoCs, it can be disabled by
hwconfig with syntax:
fsl_cpu_a011:disable
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Erratum NMG_CPU_A011 applies to P4080 rev 1.0, 2.0, fixed in rev 3.0.
It also applies to P3041 rev 1.0, 1.1, P2041 rev 1.0, 1.1. It shares the
same workaround as erratum CPU22. Rearrange registers usage in assembly
code to avoid accidental overwriting.
Signed-off-by: York Sun <yorksun@freescale.com>
Erratum A-003999: Running Floating Point instructions requires special
initialization.
Impact:
Floating point arithmetic operations may result in an incorrect value.
Workaround:
Perform a read modify write to set bit 7 to a 1 in SPR 977 before
executing any floating point arithmetic operation. This bit can be set
when setting MSR[FP], and can be cleared when clearing MSR[FP].
Alternatively, the bit can be set once at boot time, and never cleared.
There will be no performance degradation due to setting this bit.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
The P2040/P2040E have no L2 cache. So we utilize the SVR to determine
if we are one of these devices and skip the L2 init code in cpu_init.c
and release. For the device tree we skip the updating of the L2 cache
properties but we still update the chain of caches so the CPC/L3 node
can be properly updated.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
CONFIG_SYS_GBL_DATA_SIZE has always been just a bad workarond for not
being able to use "sizeof(struct global_data)" in assembler files.
Recent experience has shown that manual synchronization is not
reliable enough. This patch renames CONFIG_SYS_GBL_DATA_SIZE into
GENERATED_GBL_DATA_SIZE which gets automatically generated by the
asm-offsets tool. In the result, all definitions of this value can be
deleted from the board config files. We have to make sure that all
files that reference such data include the new <asm-offsets.h> file.
No other changes have been done yet, but it is obvious that similar
changes / simplifications can be done for other, related macro
definitions as well.
Signed-off-by: Wolfgang Denk <wd@denx.de>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
As discussed on the list, move "arch/ppc" to "arch/powerpc" to
better match the Linux directory structure.
Please note that this patch also changes the "ppc" target in
MAKEALL to "powerpc" to match this new infrastructure. But "ppc"
is kept as an alias for now, to not break compatibility with
scripts using this name.
Signed-off-by: Stefan Roese <sr@denx.de>
Acked-by: Wolfgang Denk <wd@denx.de>
Acked-by: Detlev Zundel <dzu@denx.de>
Acked-by: Kim Phillips <kim.phillips@freescale.com>
Cc: Peter Tyser <ptyser@xes-inc.com>
Cc: Anatolij Gustschin <agust@denx.de>
The HID1[MBDD] bit is new on rev5.0 or greater cores and will optimize
the performance of mbar/eieio instructions.
Signed-off-by: Sandeep Gopalpet <sandeep.kumar@freescale.com>
Use the same code between primary and secondary cores to init the
L1 cache. We were not enabling cache parity on the secondary cores.
Also, reworked the L1 cache init code to match the e500mc L2 init code
that first invalidates the cache and locks. Than enables the cache and
makes sure its enabled before continuing.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
We already map the page cache-inhibited. There is no reason we
shouldn't also be marking it guarded to prevent speculative accesses.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
The e500mc core supports the ability to stash into the L1 or L2 cache,
however we need to uniquely identify the caches with an id.
We use the following equation to set the various stash-ids:
32 + coreID*2 + 0(L1) or 1(L2)
The 0 (for L1) or 1 (for L2) matches the CT field used be various cache
control instructions.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
We need loop-check the flash clear lock and enable bit for L2 cache.
Signed-off-by: Dave Liu <daveliu@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
This change has 3 goals:
- Have secondary cores be released into spin loops at their 'true'
address in SDRAM. Previously, secondary cores were put into spin
loops in the 0xfffffxxx address range which required that boot page
translation was always enabled while cores were in their spin loops.
- Allow the TLB window that the primary core uses to access the
secondary cores boot page to be placed at any address. Previously, a
TLB window at 0xfffff000 was always used to access the seconary cores'
boot page. This TLB address requirement overlapped with other
peripherals on some boards (eg XPedite5370). By default, the boot
page TLB will still use the 0xfffffxxx address range, but this can be
overridden on a board-by-board basis by defining a custom
CONFIG_BPTR_VIRT_ADDR. Note that the TLB used to map the boot page
remains in use while U-Boot executes. Previously it was only
temporarily used, then restored to its initial value.
- Allow Boot Page Translation to be disabled on bootup. Previously,
Boot Page Translation was always left enabled after secondary cores
were brought out of reset. This caused the 0xfffffxxx address range
to somewhat "magically" be translated to an address in SDRAM. Some
boards may not want this oddity in their memory map, so defining
CONFIG_MPC8xxx_DISABLE_BPTR will turn off Boot Page Translation after
the secondary cores are initialized.
These changes are only applicable to 85xx boards with CONFIG_MP defined.
Signed-off-by: Peter Tyser <ptyser@xes-inc.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
In future Book-E implementations IVORs will most likely go away and be
replaced with fixed offsets. The IVPR will continue to exist to allow
for relocation of the interrupt vectors.
This code adds support to setup the IVORs as their fixed offset values
per the ISA 2.06 spec when we transition from u-boot to another OS
either via 'bootm' or a cpu release.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
In future Book-E implementations IVORs will most likely go away and be
replaced with fixed offsets. The IVPR will continue to exist to allow
for relocation of the interrupt vectors.
This code adds support to setup the IVORs as their fixed offset values
per the ISA 2.06 spec when we transition from u-boot to another OS
either via 'bootm' or a cpu release.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
The number of CPUs are getting detected dynamically by checking the
processor SVR value. Also removed CONFIG_NUM_CPUS references from all
the platforms with 85xx/86xx processors.
This can help to use the same u-boot image across the platforms.
Also revamped and corrected few Freescale Copyright messages.
Signed-off-by: Poonam Aggrwal <poonam.aggrwal@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Assuming the OSes exception vectors start from the base of kernel address, and
the kernel physical starting address can be relocated to an non-zero address.
This patch enables the second core to have a valid IVPR for debugger before
kernel setting IVPR in CAMP mode. Otherwise, IVPR is 0x0 and it is not a valid
value for second core which runs kernel at different physical address other
than 0x0.
Signed-off-by: Haiying Wang <Haiying.Wang@freescale.com>
Introduce CONFIG_E500MC to deal with the minor differences between
e500v2 and e500mc.
* Certain fields of HID0/1 don't exist anymore on e500mc
* Cache line size is 64-bytes on e500mc
* reset value of PIR is different
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
The e500um says the timebase is volatile out of reset. To ensure
TB sync works we need to make sure its zero.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Use CONFIG_NUM_CPUS to match existing define used by 86xx.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Acked-by: Jon Loeliger <jdl@freescale.com>
* adjust __spin_table alignment to match ePAPR v0.94 spec
* loop over all cpus when determing who is up. This fixes an issue if
the "boot cpu" isn't core0. The "boot cpu" will already be in the
cpu_up_mask so there is no harm
* Added some protection in the code to ensure proper behavior. These
changes are explicitly needed but don't hurt:
- Added eieio to ensure the "hot word" of the table is written after
all other table updates have occurred.
- Added isync to ensure we don't prefetch loading of table entries
until we a released
These issues we raised by Dave Liu.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
The following changes are needed to be inline with ePAPR v0.81:
* r4, r5 and now always set to 0 on boot release
* r7 is used to pass the size of the initial map area (IMA)
* EPAPR_MAGIC value changed for book-e processors
* changes in the spin table layout
* spin table supports a 64-bit physical release address
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Added the cpu command that provides a generic mechanism to get status,
reset, and release secondary cores in multicore processors.
Added support for using the ePAPR defined spin-table mechanism on 85xx.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Masahiro Yamada
York Sun
Wolfgang Denk
Andy Fleming
Stefan Roese
Timur Tabi
Scott Wood
Kumar Gala
Peter Tyser
Sandeep Gopalpet
Dave Liu
Poonam Aggrwal
Haiying Wang