With format-security errors turned on, GCC picks up the use of sprintf with
a format parameter not being a string literal.
Simple uses of sprintf are also converted to use strcpy.
Signed-off-by: Ben Whitten <ben.whitten@gmail.com>
Acked-by: Wolfgang Denk <wd@denx.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
We should use ARRAY_SIZE, but not sizeof. The size of
env_flags_varaccess_mask is 16bytes, but we only need 4 loops.
If using 16 as the end condition, we may access memory that
not belong to array env_flags_varaccess_mask.
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Cc: Joe Hershberger <joe.hershberger@ni.com>
Cc: York Sun <yorksun@freescale.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Report Coverity log:
Destination buffer too small (STRING_OVERFLOW)
string_overflow: You might overrun the 1024 byte destination string
lastcommand by writing 1025 bytes from console_buffer
Signed-off-by: Peng Fan <van.freenix@gmail.com>
Cc: Heiko Schocher <hs@denx.de>
Cc: Simon Glass <sjg@chromium.org>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Add a 'poweroff' command to boot commands, this only gets enabled if the
board Kconfig does a "select CMD_POWEROFF".
Signed-off-by: Michael van Slingerland <michael@deviousops.nl>
[hdegoede@redhat.com: Make the cmd conditional on a CMD_POWEROFF Kconfig]
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Extend the ums command to accept a list of block devices. Each of these
will be exported as a separate LUN. An example use-case would be:
ums 0 mmc 0,0.1,0.2
... which would export LUNs for eMMC 0's user data, boot0, and boot1 HW
partitions. This is useful since it allows the host access to everything
on the eMMC without having to somehow stop the ums command from executing
and restart it with different parameters.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
This will allow us to have multiple block device structs each referring
to the same eMMC device, yet different HW partitions.
For now, there is still a single block device per eMMC device. As before,
this block device always accesses whichever HW partition was most recently
selected. Clients wishing to make use of multiple block devices referring
to different HW partitions can simply take a copy of this block device
once it points at the correct HW partition, and use each one as they wish.
This feature will be used by the next patch.
In the future, perhaps get_device() could be enhanced to return a
dynamically allocated block device struct, to avoid the client needing to
copy it in order to maintain multiple block devices. However, this would
require all users to be updated to free those block device structs at some
point, which is rather a large change.
Most callers of mmc_switch_part() wish to permanently switch the default
MMC block device's HW partition. Enhance mmc_switch_part() so that it does
this. This removes the need for callers to do this. However,
common/env_mmc.c needs to save and restore the current HW partition. Make
it do this more explicitly.
Replace use of mmc_switch_part() with mmc_select_hwpart() in order to
remove duplicate code that skips the call if that HW partition is already
selected.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
This will allow the implementation to make use of data in the block_dev
structure beyond the base device number. This will be useful so that eMMC
block devices can encompass the HW partition ID rather than treating this
out-of-band. Equally, the existence of the priv field is crying out for
this patch to exist.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
As of gcc 5.2.1 for Thumb-1, it is not possible any
more to assign gd from C code, as gd is mapped to r9,
and r9 may now be saved in the prolog sequence, and
restored in the epilog sequence, of any C functions.
Therefore arch_setup_gd(), which is supposed to set
r9, may actually have no effect, causing U-Boot to
use a bad address to access GD.
Fix this by never calling arch_setup_gd() for ARM,
and instead setting r9 in arch/arm/lib/crt0.S, to
the value returned by board_init_f_alloc_reserve().
Signed-off-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
Reviewed-by: Simon Glass <sjg@chromium.org>
board_init_f_mem() alters the C runtime environment's
stack it is actually already using. This is not a valid
behaviour within a C runtime environment.
Split board_init_f_mem into C functions which do not alter
their own stack and always behave properly with respect to
their C runtime environment.
Signed-off-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
Acked-by: Thomas Chou <thomas@wytron.com.tw>
The function hash_show is now only called by hash_command, so mark it as
static (and drop from hash.h). We only call hash_command when any of
CONFIG_CMD_CRC32, CONFIG_CMD_SHA1SUM or CONFIG_CMD_HASH are set. Since
hash.c is linked in unconditionally we must take extra care with
functions that bring in read-only strings as these will not be
discarded.
Signed-off-by: Tom Rini <trini@konsulko.com>
To boot Linux, we should prevent Initramdisk and FDT from going too
high.
Currently, boot_relocate_fdt() checks "fdt_high" environment first,
and then falls back to getenv_bootm_mapsize() + getenv_bootm_low()
if "fdt_high" is missing.
On the other hand, boot_ramdisk_high() only checks "initrd_high" to
get the address limit for the Initramdisk. We also want to let this
case fall back to getenv_bootm_mapsize() + getenv_bootm_low().
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Adjust the SCSI command to use driver model for its PCI interface.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Tested-by: Bin Meng <bmeng.cn@gmail.com>
Most driver model PCI functions have a dm_ prefix. At some point, when the
old code is converted to driver model and the old functions are removed, we
will drop that prefix.
For consistency, we should use the dm_ prefix for all driver model
functions. Update pci_bus_find_bdf() accordingly.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Since 51209b1f42 ("Use common mtest iteration counting"),
do_mem_mtest has always reported 0 errors and hence returned 0, even
if errors were detected. Fix the helpers mem_test_alt() and
mem_test_quick() to return the number of errors found.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Commit: dm: core: Enable optional use of fdt_translate_address()
Enables use of this function as default, but after this it's not
possible to get dev address for the case in which: '#size-cells == 0'
This causes errors when getting address for some GPIOs, for which
the '#size-cells' is set to 0.
Example error:
'__of_translate_address: Bad cell count for gpx0'
Allowing for that case by modifying the macro 'OF_CHECK_COUNTS',
(called from )__of_translate_address(), fixes the issue.
Now, this macro doesn't check, that '#size-cells' is greater than 0.
This is possible from the specification point of view, but I'm not sure
that it doesn't introduce a regression for other configs.
Please test and share the results.
Tested-on: Odroid U3, Odroid X2, Odroid XU3, Sandbox.
Signed-off-by: Przemyslaw Marczak <p.marczak@samsung.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Lukasz Majewski <l.majewski@samsung.com>
Cc: Jaehoon Chung <jh80.chung@samsung.com>
Cc: Stefan Roese <sr@denx.de>
Cc: Simon Glass <sjg@chromium.org>
Cc: Bin Meng <bmeng.cn@gmail.com>
Cc: Marek Vasut <marex@denx.de>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Tested-by: Lukasz Majewski <l.majewski@samsung.com>
At present USB keyboards are not properly removed with driver model. Add the
code to handle this.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
The current limit of 5 is not enough for the driver model USB tests. Really
we should not have a limit but the driver model code still uses the
usb_dev_desc[] array, which has a limit.
Increasing the limit by 2 should not bother anyone. Adjust it.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Use snprintf to replace sprintf.
Coverity log:
"
Unbounded source buffer (STRING_SIZE)
string_size: Passing string init_val of unknown size to sprintf.
"
Reported-by: Coverity
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Cc: Tom Rini <trini@konsulko.com>
Cc: Simon Glass <sjg@chromium.org>
Reviewed-by: Joe Hershberger <joe.hershberger@ni.com>
Before calling hsearch_r, initialize callback entry to NULL.
Coverity log:
"
Uninitialized scalar variable (UNINIT)
uninit_use_in_call: Using uninitialized value e.
Field e.callback is uninitialized when calling hsearch_r.
"
Reported-by: Coverity
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Cc: Tom Rini <trini@konsulko.com>
Cc: Simon Glass <sjg@chromium.org>
Whether CONFIG_SYS_HUSH_PARSER is defined or not, should always
check to free 'buff' to avoid memory leak.
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Cc: Tom Rini <trini@konsulko.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Simon Glass <sjg@chromium.org>
Start split and complete split tokens need the hub address and the
downstream port of the first HS hub (device view).
The core of the function was duplicated in both host/ehci_hcd and
musb-new/usb-compat.h.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Marek Vasut <marex@denx.de>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
The configuration descriptor includes all interface, endpoint and
auxiliary descriptors (e.g. report, union) so 512 bytes may not be enough.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Marek Vasut <marex@denx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Currently in fdt_fixup_ethernet() the MAC address fix up is
handled in a loop of which the exit condition is to test the
"eth%daddr" env is not NULL. However this creates unnecessary
constrains that those "eth%daddr" env variables must be
sequential even if "ethernet%d" does not start from 0 in the
"/aliases" node. For example, with "/aliases" node below:
aliases {
ethernet3 = &enet3;
ethernet4 = &enet4;
};
"ethaddr", "eth1addr", "eth2addr" must exist in order to fix
up ethernet3's MAC address successfully.
Now we change the loop logic to iterate the properties in the
"/aliases" node. For each property, test if it is in a format
of "ethernet%d", then get its MAC address from corresponding
"eth%daddr" env and fix it up in the dtb.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
On OMAP4 Panda (+v4.3 kernel)
Tested-by: Tom Rini <trini@konsulko.com>
In fdt_fixup_ethernet() only "usbethaddr" is handled to fix up the
first usb ethernet port MAC address. Other additional usb ethernet
ports are ignored as there is no logic to handle "usbeth%daddr".
It is suggested we should use "ethaddr" for all ethernet devices.
Hence deprecate "usbethaddr" usage in fdt_fixup_ethernet().
This actually reverts commit b1f49ab8c7
"ARM: fdt support: Add usbethaddr as an acceptable MAC".
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
On OMAP4 Panda (+ v4.3 kernel)
Tested-by: Tom Rini <trini@konsulko.com>
When CONFIG_SYS_USB_EVENT_POLL_VIA_CONTROL_EP is enabled, use a
GET_REPORT control transfer to retrieve the initial state of the
keyboard. This matches the technique used to poll the keyboard state.
This is useful since it eliminates the remaining use of interrupt
transfers from the USB keyboard driver, which allows it to work with
USB HCD that don't support interrupt transfers.
Cc: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
This fixes commit 1a37889b0ad084a740b4f785031d7ae9955d947b:
----------------------->8--------------------
eeprom: Pull out the RW loop
Unify the code for doing read/write into single function, since the
code for both the read and write is almost identical. This again
trims down the code duplication.
----------------------->8--------------------
where the same one routine is utilized for both EEPROM writing and
reading. The only difference was supposed to be a "read" flag which
in both cases was set with 1 somehow.
That lead to a missing delay in case of writing which lead to write
failure (in my case no data was written).
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Acked-by: Marek Vasut <marex@denx.de>
Cc: Simon Glass <sjg@chromium.org>
Cc: Tom Rini <trini@konsulko.com>
Cc: Heiko Schocher <hs@denx.de>
LS1043ARDB Secure Boot Target from NOR has been added.
- Configs defined to enable esbc_validate.
- ESBC Address in header is made 64 bit.
- SMMU is re-configured in Bypass mode.
Signed-off-by: Aneesh Bansal <aneesh.bansal@freescale.com>
Reviewed-by: York Sun <yorksun@freescale.com>
As the name may be confusing, the CONFIG_SYS_MEM_TOP_HIDE reserves
some memory from the end of ram, tracked by gd->ram_size. It is not
always the top of u-boot visible memory. Rewrite the macro with a
weak function to provide flexibility for complex calcuation. Legacy
use of this macro is still supported.
Signed-off-by: York Sun <yorksun@freescale.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Secure memory is at the end of memory, separated and reserved
from OS, tracked by gd->secure_ram. Secure memory can host
MMU tables, security monitor, etc. This is different from PRAM
used to reserve private memory. PRAM offers memory at the top
of u-boot memory, not necessarily the real end of memory for
systems with very large DDR. Using the end of memory simplifies
MMU setup and avoid memory fragmentation.
"bdinfo" command shows gd->secure_ram value if this memory is
marked as secured.
Signed-off-by: York Sun <yorksun@freescale.com>
In order to support boot from multiple devices through board_boot_order,
it's necessary to use the block number of a device.
The use of a hard-coded 0 for the device number also creates a need
to re-order block devices for use in SPL like this:
http://git.denx.de/?p=u-boot.git;a=blob;f=board/freescale/mx6slevk/mx6slevk.c;hb=HEAD#l195
Signed-off-by: Eric Nelson <eric@nelint.com>
Unfortunately U-Boot assumes that almost all numbers are in hex, including
partition numbers passed to e.g. "load". So, the command "part list mmc 0
-bootable devplist" should use hex when writing partition numbers into
$devplist, so they'll be correctly interpreted.
Change-Id: I9a70b19749643876baadb45efbc3decaef8bfee2
Fixes: 0798d6fd41 ("part: Add support for list filtering on bootable partitions")
Cc: Sjoerd Simons <sjoerd.simons@collabora.co.uk>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Sjoerd Simons <sjoerd.simons@collabora.co.uk>
For a simple static string, use panic_str() which prevents calling
printf needlessly.
Signed-off-by: Sjoerd Simons <sjoerd.simons@collabora.co.uk>
Reviewed-by: Simon Glass <sjg@chromium.org>
VxWorks 7 kernels retrieve 'local-mac-addr' from dtb and use
that for NIC MAC address. As a result, when booting the same
kernel image on multiple boards, there will be address
conflicts.
So fixup MAC address when booting VxWorks 7 kernels
Signed-off-by: Miao Yan <yanmiaobest@gmail.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
With driver model timer support, there should not be an explict
call to timer_init(). Remove this call for x86.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Acked-by: Simon Glass <sjg@chromium.org>
Currently OF_BAD_ADDR is always -1ULL. When using OF_BAD_ADDR as the
return value of dev_get_addr(), it creates potential size mismatch
as dev_get_addr() uses FDT_ADDR_T_NONE as the return value which can
be either -1U or -1ULL depending on CONFIG_PHYS_64BIT. Now we change
OF_BAD_ADDR to FDT_ADDR_T_NONE to avoid such case.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Acked-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Now that we may compile (but not link) code calling fixup_cmdtable when
this is not set, we need to always have the declaration available. We
should also make sure that anyone calling the function includes
<command.h> as that's where the function declaration is.
Signed-off-by: Tom Rini <trini@konsulko.com>
Subcommands contain pointers to functions which are not updated when
MANUAL_RELOC is enabled. This patch fix it.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Acked-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Don't relocate fdt in case of CONFIG_OF EMBED as the fdt is
already embedded with u-boot image.
Update fdt_blob after relocation as the fdt will be copied
during u-boot relocation.
Signed-off-by: Siva Durga Prasad Paladugu <sivadur@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Reviewed-by: Alexey Brodkin <abrodkin@synopsys.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Tested-by: Bin Meng <bmeng.cn@gmail.com> (QEMU x86)
Tested-by: Thomas Chou <thomas@wytron.com.tw> (Nios2)
Acked-by: Thomas Chou <thomas@wytron.com.tw>
Acked-by: Simon Glass <sjg@chromium.org>
It is risky to have two different functions with much the same code. Future
authors may update one but not the other. It is hard to see which parts are
the same and which are different.
Unify the functions and drop the differences that are not really needed.
Note that one puts() becomes printf() as Tom mentioned that this does not
affect image size:
https://patchwork.ozlabs.org/patch/537276/
Note: It would be better to have an empty printf() and avoid the #ifdef for
CONFIG_SPL_LIBCOMMON_SUPPORT.
Signed-off-by: Simon Glass <sjg@chromium.org>
Tested-by: Michal Simek <michal.simek@xilinx.com>
The 'p' suffix makes it more obvious that we are dealing with a pointer
to a (pointer) value that will be returned to its caller.
Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Nikita Kiryanov <nikita@compulab.co.il>
Tested-by: Michal Simek <michal.simek@xilinx.com>
Since commit 4188ba3 we get the following warning on rockchip boards:
common/spl/spl_mmc.c:31:24: warning: ‘mmc’ may be used uninitialized in this function [-Wmaybe-uninitialized]
count = mmc->block_dev.block_read(0, sector, 1, header);
^
common/spl/spl_mmc.c:251:14: note: ‘mmc’ was declared here
struct mmc *mmc;
Correct this by move the variable init earlier.
Signed-off-by: Simon Glass <sjg@chromium.org>
Tested-by: Michal Simek <michal.simek@xilinx.com>
Acked-by: Nikita Kiryanov <nikita@compulab.co.il>
Need to free memory avoid memory leak, when error.
Signed-off-by: Peng Fan <Peng.Fan@freescale.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Cc: Simon Glass <sjg@chromium.org>
Cc: Tom Rini <trini@konsulko.com>
When CONFIG_SPL_OF_TRANSLATE is enabled fdt_support.c needs to get
compiled. Otherwise fdt_translate_address() is missing which is needed
in dev_get_addr().
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Simon Glass <sjg@chromium.org>
Acked-by: Simon Glass <sjg@chromium.org>
The following code will alloc memory for new_dev and ldev:
"
new_dev = mdio_alloc();
ldev = malloc(sizeof(*ldev));
"
Either new_dev or ldev is NULL, directly return, but this may leak memory.
So before return, using free(ldev) and mdio_free(new_dev) to avoid
leaking memory, also free can handle NULL pointer.
Signed-off-by: Peng Fan <Peng.Fan@freescale.com>
Cc: Joe Hershberger <joe.hershberger@ni.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
"enable" is unsigned char type and its value will not be
negative, so discard "enable < 0".
Signed-off-by: Peng Fan <Peng.Fan@freescale.com>
Cc: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>
Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
Cc: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Stefano Babic <sbabic@denx.de>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
The entry name of mii_dev is an array not pointer, so
no need to check.
Signed-off-by: Peng Fan <Peng.Fan@freescale.com>
Cc: Joe Hershberger <joe.hershberger@ni.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
If condition of "(load == image_start || load == image_data)" is true,
should use "fdt_addr = load;", but not "fdt_blob = (char *)image_data;",
or fdt_blob will be overridden by "fdt_blob = map_sysmem(fdt_addr, 0);"
at the end of the switch case.
Signed-off-by: Peng Fan <Peng.Fan@freescale.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Joe Hershberger <joe.hershberger@ni.com>
Cc: Max Krummenacher <max.krummenacher@toradex.com>
Cc: Marek Vasut <marex@denx.de>
Cc: Suriyan Ramasami <suriyan.r@gmail.com>
Cc: Paul Kocialkowski <contact@paulk.fr>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Condition "(value == NULL && ++value == NULL)" actully will
always return false.
Instead, use condition "(value == NULL || *(value + 1) == 0)" to detect
such expression "c=". To "c=", *(value + 1) is 0, so directly return -1,
but not continue.
Signed-off-by: Peng Fan <Peng.Fan@freescale.com>
Cc: Rabin Vincent <rabin@rab.in>
Cc: Simon Glass <sjg@chromium.org>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Fixed the prom_relocate() function in start.S file by reserving memory in
the board_init_f sequence and saving the offset to the __prom_start_reloc
variable. This value is used as the destination when relocating the PROM.
Add the prom_init() function to the end of the board_init_r sequence.
Signed-off-by: Francois Retief <fgretief@spaceteq.co.za>
Ben Whitten
Peng Fan
Peng Fan
Michael van Slingerland
Stephen Warren
Albert ARIBAUD
Tom Rini
Masahiro Yamada
Simon Glass
Rasmus Villemoes
Przemyslaw Marczak
Stefan Brüns
Bin Meng
Alexey Brodkin
Aneesh Bansal
York Sun
Eric Nelson
Sjoerd Simons
Nishanth Menon
Miao Yan
Michal Simek
Siva Durga Prasad Paladugu
Peng Fan
Stefan Roese
Francois Retief