Bin Meng [Thu, 8 Oct 2015 04:45:44 +0000 (21:45 -0700)]
net: eth: Check return value in various places
eth_get_dev() can return NULL which means device_probe() fails for
that ethernet device. Add return value check in various places or
U-Boot will crash due to NULL pointer access.
With this commit, 'dm_test_eth_act' test case passes.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Bin Meng [Thu, 8 Oct 2015 04:45:42 +0000 (21:45 -0700)]
net: eth: Clear MAC address in eth_pre_remove()
platdata->enetaddr was assigned to a value in dev_probe() last time.
If we don't clear it, for dev_probe() at the second time, dm eth
will end up treating it as a MAC address from ROM no matter where it
came from originally (maybe env, ROM, or even random). Fix this by
clearing platdata->enetaddr when removing an Ethernet device.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Bin Meng [Thu, 8 Oct 2015 04:32:39 +0000 (21:32 -0700)]
net: pch_gbe: Add driver remove support
In pch_gbe_probe(), some additional resources are allocated
(eg: mdio, phy). We should free these in the driver remove phase.
Add pch_gbe_remove() to clean it up.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Bin Meng [Thu, 8 Oct 2015 04:32:38 +0000 (21:32 -0700)]
net: designware: Add driver remove support
In designware_eth_probe(), some additional resources are allocated
(eg: mdio, phy). We should free these in the driver remove phase.
Add designware_eth_remove() to clean it up.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Acked-by: Simon Glass <sjg@chromium.org>
Bin Meng [Thu, 8 Oct 2015 04:19:31 +0000 (21:19 -0700)]
net: phy: Test previous phydev->dev against new mac dev
In phy_connect_dev(), if the phy device has an accociated mac device
before, a warning message will be printed. But we should test the
old device against the new one, if they are actually the same one,
don't print the warning message.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Bin Meng [Thu, 8 Oct 2015 04:19:30 +0000 (21:19 -0700)]
net: phy: Change to print all phys that are not found
In get_phy_device_by_mask(), when no phy is found, currently we only
print a message to show the first phy address that is not found. But
this is not always the case as multiple phys can be specified by
phy_mask. Change to print all phys that are not found, and to reduce
the console boot log, change to use 'debug' instead of 'printf'.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Sylvain Rochet [Wed, 7 Oct 2015 20:54:22 +0000 (22:54 +0200)]
net: phy: micrel: disable NAND-tree for KSZ8051
NAND-tree is used to check wiring between MAC and PHY using NAND gates
on the PHY side, hence the name.
NAND-tree initial status is latched at reset by probing the IRQ pin.
However some devices are sharing the PHY IRQ pin with other peripherals
such as Atmel SAMA5D[34]x-EK boards when using the optional TM7000
display module, therefore they are switching the PHY in NAND-tree test
mode depending on the current IRQ line status at reset.
This patch ensure PHY is not in NAND-tree test mode only for the Micrel
KSZ8051 PHY used by Atmel. There are other Micrel PHY affected but I
doubt they are used on such weird hardware design.
Signed-off-by: Sylvain Rochet <sylvain.rochet@finsecur.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stephen Warren [Fri, 2 Oct 2015 23:44:34 +0000 (17:44 -0600)]
net: rtl8169: Build warning fixes for 64-bit
Casting from dev->priv to pci_dev_t changes the value's size on a 64-bit
system. This causes the compiler to complain about casting a pointer to an
integer of a different (smaller) size. To avoid this, cast to an integer
of matching size first, then perform an int->int cast to perform the size
change. This signals explicitly that we do want to change the size, and
avoids the compiler warning. This is legitimate since we know the pointer
actually stores a small integer, not a pointer value.
Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stefan Brüns [Tue, 8 Sep 2015 03:12:00 +0000 (05:12 +0200)]
smsc95xx: Use zero length packets when RX fifo is empty
Using NAKs on empty RX fifo for bulk in transfers is the right choice
for a interrupt driven model, but U-Boot uses polling and expects an
immediate answer if there is no incoming packet. Using ZLP Bulk In Response
(BIR) mode avoids unexpected timeouts in the host controller driver.
As ZLP mode is reset default, there is no need to set it.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stefan Brüns [Thu, 3 Sep 2015 22:31:49 +0000 (00:31 +0200)]
net: parse DHCP options from overloaded file/sname fields
If Option 52 in the vendor option field signals overloading
of the file and/or sname fields, these field may contain
additional options. Formatting of file/sname contained options
is the same as in the vendor options field, but without the
leading magic.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stefan Brüns [Sun, 30 Aug 2015 15:59:45 +0000 (17:59 +0200)]
smsc95xx: Fetch whole burst with 1 URB, avoid framing errors
smsc95xx_recv() does not reassemble bursts spread over multiple URBs.
If there is a lot of broadcast traffic, the fifo will fill up to the
burst cap limit. Lowering the burst cap to the URB size ensures no packet
spans multiple urbs.
Caveat, lower limit for working burst cap is 5/33 HS/FS packets.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stefan Brüns [Sun, 30 Aug 2015 15:47:17 +0000 (17:47 +0200)]
net: cancel timeout handler after DHCPACK
Timeout handler should be stopped after reception of DHCPACK. If "autoload"
is not set, the handler is immediately replaced by the TFTP handler,
otherwise it may trigger before the next boot stage begins.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stefan Brüns [Sun, 30 Aug 2015 15:46:54 +0000 (17:46 +0200)]
net/arp: account for ARP delay, avoid duplicate packets on timeout
eth_rx() in the main reception loop may trigger sending a packet which
is already timed out (or will immediately) upon reception of an ARP reply.
As long as the ARP reply is pending, the timeout handler of a packet
should be postponed.
Happens on TFTP with bad network (e.g. WLAN).
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stefan Brüns [Thu, 27 Aug 2015 21:57:18 +0000 (23:57 +0200)]
net: send RFC1542 compliant value for bootp requests
RFC1542, 3.2:
"The 'secs' field of a BOOTREQUEST message SHOULD represent the
elapsed time, in seconds, since the client sent its first BOOTREQUEST
message. Note that this implies that the 'secs' field of the first
BOOTREQUEST message SHOULD be set to zero."
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Stefan Brüns [Thu, 27 Aug 2015 21:53:26 +0000 (23:53 +0200)]
net: reject Bootp/DHCP packets with bad OP value
Rename check_packet to check_reply_packet to make its function more
obvious.
The check for DHCP_* values is completely off, as it should
compare against DHCP option 53 (Message Type). Only valid value for
any Bootp/DHCP reply is BOOTREPLY.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
TFTP source and destination port variable names are
'tftpsrcp' and 'tftpdstp' in the code, but 'tftpsrcport'
and 'tftpdstport' in the README file. Fix the README.
Add environment variable 'tftptimeoutcountmax'. As per the
comments about the global variable tftp_timeout_count_max,
make sure tftptimeoutcountmax is nonnegative.
Introduce configuration option CONFIG_NET_TFTP_VARS,
which controls whether environment variables tftpblocksize,
tftptimeout, and tftptimoueoutcountmax are read by the TFTP
client code. CONFIG_NET_TFTP_VARS defaults to y but can be
set to n by targets with to tight size contraints.
Make bf527-ezkit set CONFIG_NET_TFTP_VARS to n to keep the
target size below limit.
Masahiro Yamada [Tue, 20 Oct 2015 12:09:06 +0000 (21:09 +0900)]
arm, powerpc: select SYS_GENERIC_BOARD
We have finished Generic Board conversion for ARM and PowerPC, i.e.
all the boards have been converted except OpenRISC, SuperH, SPARC,
which have not supported Generic Board framework yet.
Select SYS_GENERIC_BOARD in arch/Kconfig and delete all the macro
defines in include/configs/*.h.
Simon Glass [Mon, 19 Oct 2015 12:50:02 +0000 (06:50 -0600)]
zynq: Move SPL console init out of board_init_f()
We should not init the console this early since it precludes using driver
model for the UART, since it is not set up at the start of board_init_f().
See the README for more information. The debug UART does not have this
restriction. If we want to do early init with the console on it can be done
in spl_board_init().
Move the preloader_console_init() call from board_init_f() to board_init_r().
Signed-off-by: Simon Glass <sjg@chromium.org> Tested-by: Masahiro Yamada <yamada.masahiro@socionext.com> Tested-by: Michal Simek <michal.simek@xilinx.com>
Simon Glass [Mon, 19 Oct 2015 12:50:00 +0000 (06:50 -0600)]
arm: Switch 32-bit ARM to using generic global_data setup
There is quite a bit of assembler code that can be removed if we use the
generic global_data setup. Less arch-specific code makes it easier to add
new features and maintain the start-up code.
Drop the unneeded code and adjust the hooks in board_f.c to cope.
Simon Glass [Mon, 19 Oct 2015 12:49:59 +0000 (06:49 -0600)]
arm: Switch aarch64 to using generic global_data setup
There is quite a bit of assembler code that can be removed if we use the
generic global_data setup. Less arch-specific code makes it easier to add
new features and maintain the start-up code.
Drop the unneeded code and adjust the hooks in board_f.c to cope.
Tested on LS2085ARDB and LS2085AQDS (armv8 SoC). Tested-by: York Sun <yorksun@freescale.com> Signed-off-by: Simon Glass <sjg@chromium.org>
Simon Glass [Mon, 19 Oct 2015 12:49:58 +0000 (06:49 -0600)]
board_init_f_mem(): Don't create an unused early malloc() area
Change the #ifdef so that the early malloc() area is not set up in SPL if
CONFIG_SYS_SPL_MALLOC_START is defined. In that case it would never actually
be used, and just chews up stack space.
Signed-off-by: Simon Glass <sjg@chromium.org> Tested-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Simon Glass [Mon, 19 Oct 2015 12:49:56 +0000 (06:49 -0600)]
Move board_init_f_mem() into a common location
This function will be used by both SPL and U-Boot proper. So move it into
a common place. Also change the #ifdef so that the early malloc() area is
not set up in SPL if CONFIG_SYS_SPL_MALLOC_START is defined. In that case
it would never actually be used, and just chews up stack space.
Stephen Warren [Sat, 3 Oct 2015 19:56:48 +0000 (13:56 -0600)]
fs-test.sh: fix pre-requisite detection
In the following snippet:
if [ ! -x `which $prereq` ]; then
When $prereq does not exist, `which $prereq` evaluates to the empty string,
which results in *no* argument being passed to the -x operator, which then
evaluates to true, which is the equivalent of the prereq having been found. In
order for this to fail as expected, we must pass an empty argument, which then
causes -x to fail. Do this by wrapping the `` in quotes so there's always an
argument to -x, even if the value of the argument is zero-length.
Signed-off-by: Stephen Warren <swarren@wwwdotorg.org> Reviewed-by: Simon Glass <sjg@chromium.org> Tested-by: Simon Glass <sjg@chromium.org>
Stephen Warren [Sat, 3 Oct 2015 19:56:47 +0000 (13:56 -0600)]
test: fat: add test of non-contiguous file reads
In my patch series to replace fs/fat with "ff.c", I enhanced ff.c to
optimize file reading, so that reads of contiguous clusters are submitted
to the IO device as a single read. This test attempts to torture-test
edge-cases of that enhancement.
BTW, the only way I found to validate that this script actually does
create non-contiguous files was to manually inspect the FAT bitmap in a
hex dump of the FAT image. hdparm --fibmap doesn't work on loop-mounted
filesystems. filefrag -v -e seems to lie about files being contiguous
when they aren't.
Signed-off-by: Stephen Warren <swarren@wwwdotorg.org> Reviewed-by: Simon Glass <sjg@chromium.org> Tested-by: Simon Glass <sjg@chromium.org>
Hans de Goede [Thu, 17 Sep 2015 22:46:55 +0000 (18:46 -0400)]
disk/part: Only build hostfs special handling when CONFIG_SANDBOX is set
This is not necessary / useful when not building with CONFIG_SANDBOX and
with the addition of ubifs support to the generic fs commands it actually
gets in the way, since both operate on a fake / NULL blkdev.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Acked-by: Stephen Warren <swarren@nvidia.com>
Thomas Huth [Tue, 25 Aug 2015 15:09:40 +0000 (17:09 +0200)]
Fix bad return value checks (detected with Coccinelle)
In the "Getting Started with Coccinelle - KVM edition" presentation that
has been held by Julia Lawall at the KVM forum 2015 (see the slides at
http://events.linuxfoundation.org/sites/events/files/slides/tutorial_kvm_0.pdf),
she pointed out some bad return value checks in U-Boot that can be
detected with Coccinelle by using the following config file:
@@
identifier x,y;
identifier f;
statement S;
@@
x = f(...);
(
if (x < 0) S
|
if (
- y
+ x
< 0) S
)
Guillaume GARDET [Tue, 25 Aug 2015 13:10:26 +0000 (15:10 +0200)]
ARM: rpi: Add CONFIG_ENV_VARS_UBOOT_RUNTIME_CONFIG support
Add CONFIG_ENV_VARS_UBOOT_RUNTIME_CONFIG support and enable it to set
'board_rev' and 'board_name' envs.
'board_rev' can be used in scripts to determine what board we are running on
and 'board_name' for pretty printing.
Signed-off-by: Guillaume GARDET <guillaume.gardet@free.fr> Cc: Stephen Warren <swarren@wwwdotorg.org> Acked-by: Stephen Warren <swarren@wwwdotorg.org>
For current U-Boot to initialize status LEDs via status_led_init(), it
is required to have both CONFIG_STATUS_LED and STATUS_LED_BOOT defined.
This may be a particular concern with GPIO LEDs, where __led_init() is
required to correctly set up the GPIO (gpio_request and
gpio_direction_output). Without STATUS_LED_BOOT the initialization isn't
called, which could leave the user with a non-functional "led" command -
due to the fact that the LED routines in gpio_led.c use gpio_set_value()
just fine, but the GPIO never got set up properly in the first place.
I think having CONFIG_STATUS_LED is sufficient to justify a
corresponding call to status_led_init(), even with no STATUS_LED_BOOT
defined. To do so, common/board_r.c needs call that routine, so it now
is exposed via status_led.h.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de>
[trini: Add dummy __led_init to pca9551_led.c] Signed-off-by: Tom Rini <trini@konsulko.com>
For boards that support LEDs driven via GPIO (CONFIG_GPIO_LED),
it may be useful to have some generic stubs (wrapper functions)
for the "colored" LEDs.
This allows defining STATUS_LED_* values directly to GPIO numbers,
e.g.: #define STATUS_LED_GREEN 248 /* = PH24 */
To keep those optional, it's probably best to introduce an additional
configuration setting. I've chosen CONFIG_GPIO_LED_STUBS for that.
Placing the code in drivers/misc/gpio_led.c also ensures that it
automatically depends on CONFIG_GPIO_LED too.
Signed-off-by: Bernhard Nortmann <bernhard.nortmann@web.de> Reviewed-by: Simon Glass <sjg@chromium.org>
Thomas Chou [Thu, 22 Oct 2015 23:58:20 +0000 (07:58 +0800)]
nios2: convert cache flush to use dm cpu data
Convert cache flush to use dm cpu data.
The original cache flush functions are written in assembly
and use CONFIG_SYS_{I,D}CACHE_SIZE... macros. It is difficult
to convert to use cache configuration in dm cpu data which is
extracted from device tree.
The cacheflush.c of Linux nios2 arch uses cpuinfo structure,
which is very close to our dm cpu data. So we copy and modify
it to arch/nios2/lib/cache.c to replace the old cache.S.
Thomas Chou [Fri, 9 Oct 2015 12:09:17 +0000 (20:09 +0800)]
nios2: set default cache configuration in start.S
Set default icache and dcache configuration for start.S.
We want to remove the CONFIG_SYS_{I,D}CACHE_SIZE...
configuration macros. As we are just barely starting from
reset, there is no luxury of device tree.
We will set some maximum cache configuration so that it will
work for most configurations. This is used only in this
start.S. The speed penalty is only once here.
After start up, during board initialization, cpu information
will be extracted from device tree. Then cache flush operations
will have correct cache configurations.
Signed-off-by: Thomas Chou <thomas@wytron.com.tw> Acked-by: Marek Vasut <marex@denx.de>
Thomas Chou [Wed, 21 Oct 2015 13:34:57 +0000 (21:34 +0800)]
nios2: convert nios2 cpu to driver model
Convert nios2 cpu to driver model. The cpu parameters are
extracted from device tree and saved to global data structure.
We will use them to replace the custom_fpga.h .
Signed-off-by: Thomas Chou <thomas@wytron.com.tw> Reviewed-by: Simon Glass <sjg@chromium.org>
Thomas Chou [Wed, 9 Sep 2015 07:09:43 +0000 (15:09 +0800)]
nios2: call board_init_f_mem
We will need CONFIG_SYS_MALLOC_F_LEN to use serial uclass.
So we shall undefine CONFIG_SYS_GENERIC_GLOBAL_DATA, and
call board_init_f_mem() to allocates early malloc() memory
with size of CONFIG_SYS_MALLOC_F_LEN in board_f.c.
Signed-off-by: Thomas Chou <thomas@wytron.com.tw> Reviewed-by: Marek Vasut <marex@denx.de>
Thomas Chou [Sun, 6 Sep 2015 12:13:34 +0000 (20:13 +0800)]
nios2: define _end in link script
Since commit 44c6e6591cb451ae606f8bde71dd5fb7b4002544
"rename _end to __bss_end__" , the _end was removed.
But we need it now for separated device tree control,
ie, CONFIG_OF_SEPARATE .
The _end is used by fdtdec_setup() to find the blob.
Signed-off-by: Thomas Chou <thomas@wytron.com.tw> Acked-by: Marek Vasut <marex@denx.de> Reviewed-by: Simon Glass <sjg@chromium.org>
Thomas Chou [Fri, 4 Sep 2015 08:39:16 +0000 (16:39 +0800)]
nios2: enlarge the code relocation range
As we will use u-boot-dtb.bin, the code relocation range
should be adjusted to accommodate the additional dtb.
It might be overkilled to look into dtb header to find the
dtb size, so we will simply use CONFIG_SYS_MONITOR_LEN.
Mugunthan V N [Tue, 13 Oct 2015 08:32:29 +0000 (14:02 +0530)]
ARM: AM335x: mux: change mmc0 cd pinmux from mmc0_sdcd to gpio
Currently omap_hsmmc driver doesn't use sdcd pin to detect
whether the card is present or not. Instead the same pin is used
as GPIO to detect card presence. So change the pin mux mode from
mmc0_sdcd to gpio0_6.
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com>
Mugunthan V N [Tue, 13 Oct 2015 08:27:16 +0000 (13:57 +0530)]
drivers: gpio: omap: add support for parsing additional gpio parameters
With DM_GPIO, gpio parameters like ACTIVE_(LOW/HIGH) are to be
parsed in xlate gpio drivers-ops. Since xlate is not implemented
in omap_gpio driver, the driver considers all gpio to be
ACTIVE_HIGH which is the default case and fails to return actual
gpio status for ACTIVE_LOW gpios. So adding .xlate ops to
omap_gpio.
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com> Reviewed-by: Simon Glass <sjg@chromium.org>
projects / u-boot / log