- 'pciconfighost' is set by default in order to be able to scan bridges
behind the primary host/PCIe
- 'pciscandelay' env variable is recognized to allow for user-controlled
delay before the PCIe bus enumeration; some peripheral devices require a
significant delay before they can be scanned (e.g. LSI8408E); without the
delay they are not detected
Signed-off-by: Grzegorz Bernacki <gjb@semihalf.com>
- correct configuration space mapping
- correct bus numbering
- better access to config space
Prior to this patch, the 440SPe host/PCIe bridge was able to configure only the
first device on the first bus. We now allow to configure up to 16 buses;
also, scanning for devices behind the PCIe-PCIe bridge is supported, so
peripheral devices farther in hierarchy can be identified.
Signed-off-by: Grzegorz Bernacki <gjb@semihalf.com>
Convert using fixup mechanism to suppressing MCK for the duration of config
read/write transaction: while fixups work fine with the case of a precise
exception, we identified a major drawback with this approach when there's
an imprecise case. In this scenario there is the following race condition:
the fixup is (by design) set to catch the instruction following the one
actually causing the exception; if an interrupt (e.g. decrementer) happens
between those two instructions, the ISR code is executed before the fixup
handler the machine check is no longer protected by the fixup handler as it
appears as within the ISR code. In consequence the fixup approach is being
phased out and replaced with explicit suppressing of MCK during a PCIe
config read/write cycle.
Signed-off-by: Grzegorz Bernacki <gjb@semihalf.com>
When frame larger than local RX buffer is received, it is split and handled
by two buffer descriptors. Prior to this patch the FEC driver discarded
contents of a buffer descriptor without the 'LAST' bit set, so the first
part of the frame was lost in case of larger frames. This fix allows to
safely combine the two pieces into the whole frame.
Signed-off-by: Grzegorz Bernacki <gjb@semihalf.com>
This patch adds support for multiple serial ports to the PXA target.
FFUART, BTUART and STUART are supported.
Signed-off-by: Stefano Babic <sbabic@denx.de>
Stop using inline string functions on MIPS as other ARCHs do so,
since the optimized inline asm versions are not small.
This change is triggered by a following MIPS build error:
common/libcommon.a(exports.o)(.text+0xdc): In function `jumptable_init':
common/exports.c:32: undefined reference to `strcmp'
make: *** [u-boot] Error 1
Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi@necel.com>
This patches contains several bugfixes and cleanups in the latest upstream:
- Don't include linux/config.h
- Remove buggy inline version of memscan.
- Merge with Linux 2.6.11-rc3.
- Fix undefined reference to strcpy in binfmt_misc caused by gcc 3.4.
- Goodbye mips64. 31704 lines of code bite the dust.
- Replace extern inline with static inline.
- Fix return value of strncpy.
- Remove a bunch more "$1" clobbers.
Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi@necel.com>
Fix incorrect blob address calculation in do_bootm_linux() that prevents
booting the kernel from a multi-file image (kernel + initrd + blob).
Also, make minor updates to the U-Boot's output and to the coding style.
Signed-off-by: Bartlomiej Sieka <tur@semihalf.com>
Boards with CONFIG_OF_LIBFDT enabled are not able to boot old-style
kernels using the board info structure (instead of passing a device tree)
This change allows the old style booting to be used if the fdt argument
was not passed to 'bootm'.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Acked-by: Kim Phillips <kim.phillips@freescale.com>
Affects boards: icecube (lite5200), jupiter, motionpro, tqm5200
Tested on: lite5200b
Note: the fixup functions have not been moved to a common place. This
patch is targeted for immediate merging as in solves a build issue, but
the final name/location of the fixups is still subject to debate. I
propose to merge this now, and move the fixups in the next merge window
to be usable by all targets.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
The RTC on the M1575 ULI chipset requires a dummy read before
we are able to talk to the RTC. We accomplish this by adding a
second memory region to the PHB the ULI is on and read from it.
The second region is added to maintain compatiabilty with Linux's
view of the PCI memory map.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
The 440EPx has a problem when the PCI_CACHE_LINE_SIZE register is
set to non-zero, because it doesn't support MRM (memory-read-
multiple) correctly. We now added the possibility to configure
this register in the board config file, so that the default value
of 8 can be overridden.
Here the details of this patch:
o drivers_pci_auto.c: introduce CFG_PCI_CACHE_LINE_SIZE to allow
board-specific settings. As an example the sequoia board requires 0.
Idea from Stefan Roese <sr@denx.de>.
o board/amcc/sequoia/init.S: add a TLB mapping at 0xE8000000 for the
PCI IO-space. Obtained from Stefan Roese <sr@denx.de>.
o include/configs/sequoia.h: turn CONFIG_PCI_PNP back on and set
CFG_PCI_CACHE_LINE_SIZE to 0.
Signed-off-by: Gary Jennejohn <garyj@denx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
The old code had a loop limit overflow bug which caused a semi-
infinite loop for small packets, because in "i<len-8", "i" was signed,
but "len" was unsigned, and "len-8" became a huge number for small
values of "len".
This is a workaround which replaces broken commit 8f1bc284.
Signed-off-by: Wolfgang Denk <wd@denx.de>
The TFTP OACK code trusts that the incoming packet is formated as
ASCII text and can be processed by string functions. It also has a
loop limit overflow bug where if the packet length is less than 8, it
ends up looping over *all* of memory to find the 'blksize' string.
This patch solves the problem by forcing the packet to be null
terminated and using strstr() to search for the sub string.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
The Linux PCF8563 RTC driver doesn't do autoprobing, so we need
to supply the RTC I2C address as bootline parameter. This patch
adds support for this rtc probing parameter to the bootargs:
"rtc-pcf8563.probe=0,0x51"
Signed-off-by: Stefan Roese <sr@denx.de>
this had slipped through the cracks, since the sbc board was added
after I wrote the original patch to remove all these symbols, and
before it was merged.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
<ed.swarthout@freescale.com>
The problem is pciauto_setup_device() getting called from fsl_pci_init.c
is allocating memory space it doesn't need.
Signed-off-by: Ed Swarthout <ed.swarthout@freescale.com>
Signed-off-by: Andy Fleming <afleming@freescale.com>
Restore CONFIG_EXTRA_ENV_SETTINGS definition which contains the
correct consoledev needed for linux boot.
Standardize on fdt{file,addr} var to hold dtb file name.
Set PCI inbound memory region from CFG_MEMORY_{BUS,PHYS}.
Signed-off-by: Ed Swarthout <Ed.Swarthout@freescale.com>
if a board/$(VENDOR)/common/Makefile exists, build it.
also add the first such case, board/freescale/common/Makefile, to
handle building board-shared EEPROM, PIXIS, and MDS-PIB code, as
dictated by board configuration.
thusly get rid of alternate build dir errors such as:
FATAL: can't create /work/wd/tmp/u-boot-ppc/board/freescale/mpc8360emds/../common/pq-mds-pib.o: No such file or directory
by putting the common/ mkdir command in its proper place (the common
Makefile). Common bits from existing individual board Makefiles have
been removed.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
TFTP does not work anymore after multicast tftp
patch was applied on little endian systems.
This patch fix it.
Signed-off-by: Stefano Babic <sbabic@denx.de>
The logic to check if there is a correct MAC address in the DM9000
EEPROM, added in the last patch, is wrong. Now the MAC address is
always taken from the environment, even if a suitable MAC is present
in the EEPROM.
Signed-off-by: Stefano Babic <sbabic@denx.de>
Fix the following warnings:
- usb.c:xx: warning: function declaration isn't a prototype
- usb_ohci.c:xxx: warning: passing argument 1 of '__fswab32' makes integer
from pointer wihtout a cast
Signed-off-by: Martin Krause <martin.krase@tqs.de>
CPU physical address space was being wasted by allocating a
PCSRBAR PCI inbound region to it's memory space.
As a rule, PCSRBAR should be left alone since it does not affect
transactions from self and other masters may have changed it.
Signed-off-by: Ed Swarthout <ed.swarthout@freescale.com>
