In the 36-bit physical config for MPC8572DS when need the start address
of memory and it size to be kept in phys_*_t instead of a ulong since
we support >4G of memory in the config and ulong cant represent that.
Otherwise we end up seeing the memory node in the device tree reporting
back we have memory starting @ 0 and of size 0.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
When we introduced the 36-bit config of the MPC8572DS board we had the
wrong PCI MEM bus address map. Additionally, the change to the address
map exposes a small issue in our dummy read on the ULI bus. We need
to use the new mapping functions to handle that read properly in the
36-bit config.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Previously we only allowed power-of-two memory sizes and didnt
handle >2G of memory. Now we will map up to CONFIG_MAX_MEM_MAPPED
and should properly handle any size that we can make in the TLBs
we have available to us
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
If we only have one controller we can completely ignore how
memctl_intlv_ctl is set. Otherwise other levels of code get confused
and think we have twice as much memory.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
The LUN number is not part of the Command Descriptor Block (CDB) for scsi inquiry, request sense, test unit ready, read capacity and read10 commands. This patch removes the LUN number information from the CDB.
Signed-off-by: Thomas Abraham <t-abraham@ti.com>
Signed-off-by: Remy Bohmer <linux@bohmer.net>
V3: Fixed line-wrap problem due to user error in mail!
Added usb_configured() checks in usbtty_puts() and usbtty_putc() to get around a hang
when usb is not connected and the user has set up multi-io (setenv stdout serial,usbtty etc).
Got rid of redundant __attribute__((packed)) directives that were causing warnings from gcc.
Signed-off-by: Atin Malaviya <atin.malaviya@gmail.com>
Signed-off-by: Remy Bohmer <linux@bohmer.net>
i.MX31 powers on with most clocks running, so, after a power on this explicit
clock start up is not required. However, as Linux boots it disables most clocks
to save power. This includes the I2C clock. If we then soft reboot from Linux
the I2C clock stays off. This breaks the phycore, which has its environment in
I2C EEPROM. Fix the problem by explicitly starting the clock in I2C driver
initialisation routine.
Signed-off-by: Guennadi Liakhovetski <lg@denx.de>
Ack-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
The Blackfin i2c driver has been rewritten thus the special ifdefs in the
common code are no longer needed.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
With actual u-boot compiling the mgcoge port fails, because
since commit ba705b5b1a it is
necessary to define CONFIG_NET_MULTI.
Seems to me the mgcoge port is the only actual existing 8260
port who uses CONFIG_ETHER_ON_SCC, so no other 8260 port needed
to be fixed.
Signed-off-by: Heiko Schocher <hs@denx.de>
This fixes the initialization of the SDRAM_CODT register in the ppc4xx DDR2
initialization code. It also removes use of the SDRAM_CODT_FEEDBACK_RCV_SINGLE_END
and SDRAM_CODT_FEEDBACK_DRV_SINGLE_END #define's since they are reserved bits.
Signed-off-by: Carolyn Smith <carolyn.smith@tektronix.com>
Signed-off-by: Stefan Roese <sr@denx.de>
Some AMCC eval boards do have a board_eth_init() function calling
pci_eth_init(). These boards need to call cpu_eth_init() explicitly now
with the new eth_init rework.
Signed-off-by: Stefan Roese <sr@denx.de>
The criteria of the AMCC SDRAM Controller DDR autocalibration
U-Boot code is to pick the largest passing write/read/compare
window that also has the smallest SDRAM_RDCC.[RDSS] Read Sample
Cycle Select value.
On some Kilauea boards the DDR autocalibration algorithm can
find a large passing write/read/compare window with a small
SDRAM_RDCC.[RDSS] aggressive value of Read Sample Cycle Select
value "T1 Sample".
This SDRAM_RDCC.[RDSS] Read Sample Cycle Select value of
"T1 Sample" proves to be to aggressive when later on U-Boot
relocates into DDR memory and executes.
The memory traces on the Kilauea board are short so on some
Kilauea boards the SDRAM_RDCC.[RDSS] Read Sample Cycle Select
value of "T1 Sample" shows up as a potentially valid value for
the DDR autocalibratiion algorithm.
The fix is to define a weak default function which provides
the minimum SDRAM_RDCC.[RDSS] Read Sample Cycle Select value
to accept for DDR autocalibration. The default will be the
"T2 Sample" value. A board developer who has a well defined
board and chooses to be more aggressive can always provide
their own board specific string function with the more
aggressive "T1 Sample" value or stick with the default
minimum SDRAM_RDCC.[RDSS] value of "T2".
Also put in a autocalibration loop fix for case where current
write/read/compare passing window size is the same as a prior
window size, then in this case choose the write/read/compare
result that has the associated smallest RDCC T-Sample value.
Signed-off-by: Adam Graham <agraham@amcc.com>
Signed-off-by: Stefan Roese <sr@denx.de>
CONFIG_SDRAM_PPC4xx_IBM_DDR2 is not set when include/asm-ppc/config.h is
included. So for katmai, CONFIG_MAX_MEM_MAPPED will get set to 256MB.
It makes perfect sense to set CONFIG_MAX_MEM_MAPPED to 2GB for all PPC4xx
boards right now.
Signed-off-by: Stefan Roese <sr@denx.de>
This patch adds the configuration option CONFIG_SYS_SMC_RXBUFLEN.
With this option it is possible to allow the receive
buffer for the SMC on 8xx to be greater then 1. In case
CONFIG_SYS_SMC_RXBUFLEN == 1 this driver works as the
old version.
When defining CONFIG_SYS_SMC_RXBUFLEN also
CONFIG_SYS_MAXIDLE must be defined to setup the maximum
idle timeout for the SMC.
Signed-off-by: Heiko Schocher <hs@denx.de>
Move global '#ifdef CONFIG_xxx .... #endif' out of the .c files and into
the COBJS-$(CONFIG_xxx) in the Makefile. Also delete unused var in kgdb
code in the process.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
At some point an intentional double space at the end of the sentence
got changed into a tab in the GPL header line:
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
This patch fixes the damage.
Signed-off-by: Gerald Van Baren <vanbaren@cideas.com>
If on your board is more than one flash, you must know
the size of every single flash, for example, for updating
the DTS before booting Linux. So make this function
flash_get_info() extern, and you can have all info
about your flashes.
Signed-off-by: Heiko Schocher <hs@denx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
Added new CONFIG options for the three type of MAC-PHY interconnect and
applied them all relevant board config files
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
SOFT_RESET must be asserted for at least 3 TX clocks. Usually, that's about 30
clock cycles, so it's been mostly working. But we had no guarantee, and at
slower bitrates, it's just over a microsecond (over 1000 clock cycles). This
enforces a 2 microsecond gap between assertion and deassertion.
Signed-off-by: Andy Fleming <afleming@freescale.com>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
100Mbs ethernet does not work on sh7763 chips due to the wrong value being
used in the GECMR register. Following diff fixes the problem
Signed-off-by: Simon Munton <simon@nidoran.m5data.com>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
This fixes MPC8260 compilation with ethernet on SCC. Probably was a
typo or something...
Signed-off-by: Sergey Kubushyn <ksi@koi8.net>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
This patch adds the configuration option CONFIG_SYS_SMC_RXBUFLEN.
With this option it is possible to allow the receive
buffer for the SMC on 82xx to be greater then 1. In case
CONFIG_SYS_SMC_RXBUFLEN == 1 this driver works as the
old version.
When defining CONFIG_SYS_SMC_RXBUFLEN also
CONFIG_SYS_MAXIDLE must be defined to setup the maximum
idle timeout for the SMC.
Signed-off-by: Heiko Schocher <hs@denx.de>
If we call flush_cache(0xfffff000, 0x1000) it would never
terminate the loop since end = 0xffffffff and we'd roll over
our counter from 0xfffffe0 to 0 (assuming a 32-byte cache line)
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Moved CONFIG_MAX_MEM_MAPPED to the asm/config.h so its kept consistent
between the two current users (lib_ppc/board.c, 44x SPD DDR2).
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Acked-by: Stefan Roese <sr@denx.de>
We have common defines that we duplicate in various ways. Having an
arch specific config.h gives us a common location for those defines.
Eventually we should be able to replace this when we have proper
Kconfig support.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Now that the rest of u-boot can support it, change the PCI bus
address of the PCI MEM regions from 0x80000000 to 0xc0000000,
and use the same bus address for both PCI1 and PCI2. This will
maximize the amount of PCI address space left over to map RAM
on systems with large amounts of memory.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
The code assumes that the pci bus address and the virtual
address used to access a region are the same, but they might
not be. Fix this assumption.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
Clean up PCI mapping concepts in the 8641 config - rename _BASE
to _BUS, as it's actually a PCI bus address, separate virtual
and physical addresses into _VIRT and _PHYS, and use each
appopriately.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
The BAT fields are architected; there's no need for these to be in
cpu-specific files. Drop the duplication and move these to
include/asm-ppc/mmu.h. Also, remove the BL_xxx defines that were only
used by the alaska board, and switch to using the BATU_BL_xxx defines
used by all the other boards. The BL_ defines previously in use
had to be shifted into the proper position for use, which was inefficient.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
It is no longer always true that the pci bus address can be
used as the virtual address for pci accesses. pci_map_bar()
is created to return the virtual address for a pci region.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
Because the inbound pci windows are mapped generously, set up
the more specific outbound windows first. This way, when we
search the pci regions for something, we will hit on the more
specific region. This can actually be a problem on systems
with large amounts of RAM.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
If the VA and PA of the flash aren't the same, the banks list
should be initialized to hold the physical address. Correct this.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
Driver for Dave DNET ethernet controller (used on Dave/DENX
QongEVB-LITE board).
Signed-off-by: Ilya Yanok <yanok@emcraft.com>
Acked-by: Ben Warren <biggerbadderben@gmail.com>
When we search for an address match in pci_hose_{phys_to_bus,bus_to_phys}
we should give preference to memory regions that aren't system memory.
Its possible that we have over mapped system memory in the regions and
we want to avoid depending on the order of the regions.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
