One specific USB 3.0 device behaves strangely when reset by
usb_new_device()'s call to hub_port_reset(). For some reason, the device
appears to briefly drop off the bus when this second bus reset is
executed, yet if we retry this loop, it'll eventually come back after
another two resets.
If USB bus reset is executed over and over within usb_new_device()'s call
to hub_port_reset(), I see the following sequence of results, which
repeats as long as you want:
1) STAT_C_CONNECTION = 1 STAT_CONNECTION = 0 USB_PORT_STAT_ENABLE 0
2) STAT_C_CONNECTION = 1 STAT_CONNECTION = 1 USB_PORT_STAT_ENABLE 0
3) STAT_C_CONNECTION = 1 STAT_CONNECTION = 1 USB_PORT_STAT_ENABLE 1
The device in question is a SanDisk Ultra USB 3.0 16GB memory stick with
USB VID/PID 0x0781/0x5581.
In order to allow this device to work with U-Boot, ignore the
{C_,}CONNECTION bits in the status/change registers, and only use the
ENABLE bit to determine if the reset was successful.
To be honest, extensive investigation has failed to determine why this
problem occurs. I'd love to know! I don't know if it's caused by:
* A HW bug in the device
* A HW bug in the Tegra USB controller
* A SW bug in the U-Boot Tegra USB driver
* A SW bug in the U-Boot USB core
This issue only occurs when the device's USB3 pins are attached to the
host; if only the USB2 pins are connected the issue does not occur. The
USB3 controller on Tegra is in reset, so is not actively communicating
with the device at all - a USB3 analyzer confirms this. Slightly
unplugging the device (so the USB3 pins don't contact) or using a USB2
cable or hub as an intermediary avoids the problem. For some reason,
the Linux kernel (either on the same Tegra board, or on an x86 host)
has no issue with the device, and I observe no disconnections during
reset.
This change won't affect any USB device that already works, since such
devices could not currently be triggering the error return this patch
removes, or they wouldn't be working currently.
However, this patch is quite reliable in practice, hence I hope it's
acceptable to solve the problem.
The only potential fallout I can see from this patch is:
* A broken device that triggers C_CONNECTION/!CONNECTION now causes the
loop in hub_port_reset() to run multiple times. If it never succeeds,
this will cause "usb start" to take roughly 1s extra to execute.
* If the user unplugs a device while hub_port_reset() is executing, and
very quickly swaps in a new device, hub_port_reset() might succeed on
the new device. This would mean that any information cached about the
original device (from the descriptor read in usb_new_device(), which
simply caches the max packet size) might be invalid, which would cause
problems talking to the new device. However, without this change, the
new device wouldn't work anyway, so this is probably not much of a
loss.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
As we support both Host and Device mode operation, an OTG controller
can return -ENODEV on a port which it found to be in Device mode during
Host mode scan for devices. In case -ENODEV is returned, print that the
port is not available and continue instead of screaming a bloody error
message.
Signed-off-by: Marek Vasut <marex@denx.de>
Commit 3ff291f371
(kconfig: convert Kconfig helper script into a shell script)
restored "<board>_config" target for backward compatibility.
It should be documented.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
This tool only works on python 2 (python 2.6 or lator).
Change the shebang to make sure the script is run by python 2
and clearly say the supported version in the comment block.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
We noticed on the DXR2 platform (AM335x with a SMSC LAN9303 switch connected
to the CPSW MAC) that the network performance in U-Boot is quite poor. Only
when the transfer is started without a cable connected, and the cable is
plugged after the first timeout "T" occured, an increased in performance
can be seen. Debugging has revealed, that the cpsw driver has constant
link checking builtin into the rx and tx functions. This results in the
bad performance and seems to be unnecessary. The link has already been
checked in the init function, before the transfer is started. This usually
is sufficient.
BTW: I have seen no other network driver in U-Boot so far, that constantly
checks for link in the rx / tx functions.
The performance numbers on the DXR2 board are:
0.56 MiB/s cpsw_check_link() in rx and tx path
0.87 MiB/s cpsw_check_link() only in tx path
1.0 MiB/s cpsw_check_link() only in rx path
2.7 MiB/s no cpsw_check_link() in rx and tx path
So with this patch the network performance on DXR2 increases from 0.56
to 2.7 MiB/s (nearly 5 times as fast).
Signed-off-by: Stefan Roese <sr@denx.de>
Tested-by: Samuel Egli <samuel.egli@siemens.com>
Tested-by: Heiko Schocher <hs@denx.de>
Cc: Vladimir Koutny <vladimir.koutny@streamunlimited.com>
Cc: Mugunthan V N <mugunthanvnm@ti.com>
Cc: Joe Hershberger <joe.hershberger@gmail.com>
Cc: Tom Rini <trini@ti.com>
I guess some developers are already getting sick of this tool
because it generally takes a few minites to generate the boards.cfg
on a reasonable computer.
The idea popped up on my mind was to skip Makefiles and
to run script/kconfig/conf directly.
This tool should become about 4 times faster.
You might still not be satisfied, but better than doing nothing.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
It looks silly to regenerate the boards.cfg even when it is
already up to date.
The tool should exit with doing nothing if the boards.cfg is newer
than any of defconfig, Kconfig and MAINTAINERS files.
Specify -f (--force) option to get the boards.cfg regenerated
regardless its time stamp.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
This tool deletes the incomplete boards.cfg
if it encounters an error or is is terminated by the user.
I notice some problems even though they rarely happen.
[1] The boards.cfg is removed if the program is terminated
during __gen_boards_cfg() function but before boards.cfg
is actually touched. In this case, the previous boards.cfg
should be kept as it is.
[2] If an error occurs while deleting the incomplete boards.cfg,
the program throws another exception. This hides the privious
exception and we will not be able to know the real cause.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
When an error occurs or the program is terminated by the user
on the way, the destructer __del__ of class Slot is invoked and
the work directories are removed.
We have to make sure there are no subprocesses (in this case,
"make O=<work_dir> ...") using the work directories before
removing them. Otherwise the subprocess spits a bunch of error
messages possibly causing more problems. Perhaps some users
may get upset to see too many error messages.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
The tools/genboardscfg.py expects all the Kconfig and defconfig are
written correctly. Imagine someone accidentally has broken a board.
Error-out just for one broken board is annoying for the other
developers. Let the tool skip insane boards and continue processing.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
tools/genboardscfg.py expects all the boards have MAINTAINERS.
If someone adds a new board but misses to add its MAINTAINERS file,
tools/genboardscfg.py fails to generate the boards.cfg file.
It is annoying for the other developers.
This commit allows tools/genboardscfg.py to display warning messages
and continue processing even if some MAINTAINERS files are missing
or have broken formats.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
Kconfig in U-Boot creates a temporary file configs/.tmp_defconfig
during processing "make <board>_defconfig". The temporary file
might be left over for some reasons.
Just in case, tools/genboardscfg.py should make sure to
not read such garbage files.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
The host filesystem name has changed, so update the tests. The tests now
run again correctly:
$ make O=b/sandbox sandbox_defconfig all
...
$ test/image/test-fit.py -u b/sandbox/u-boot
FIT Tests
=========
Kernel load
Kernel + FDT load
Kernel + FDT + Ramdisk load
Tests passed
Caveat: this is only a sanity check - test coverage is poor
Signed-off-by: Simon Glass <sjg@chromium.org>
Commit e3a5bbce broke the FIT image tests by not loading a ramdisk even if
a load address is provided in the FIT. The rationale was that a load address
of 0 should be considered to mean 'do not load'.
Add a new load operation which supports this feature, so that the ramdisk
will be loaded if a non-zero load address is provided.
Signed-off-by: Simon Glass <sjg@chromium.org>
Copy the Kconfig option from "init/Kconfig" of Linux v3.16 tag
and adjust the help document.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Now we have CONFIG_LOCALVERSION and CONFIG_LOCALVERSION_AUTO
in Kconfig so we can use scripts/setlocalversion without
any adjustment. Copy it from Linux 3.16 as is.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Copy Kconfig options from "init/Kconfig" of Linux v3.16 tag
and adjust some parts of the help document.
Move CONFIG_SPL, CONFIG_TPL, ... etc. to "Boot images" menu.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
This commit makes sure boards.cfg is up to date before starting
the build tests. tools/genboardscfg.py exits immediately printing
"boards.cfg is up to date. Nothing to do." when boards.cfg is
already new.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Simon Glass <sjg@chromium.org>
This commit makes sure boards.cfg is up to date before starting
the build tests. tools/genboardscfg.py exits immediately printing
"boards.cfg is up to date. Nothing to do." when boards.cfg is
already new.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
This patch fixes a minor problem:
If a block without "F: configs/*_defconfig" is followed by another
block with "F: configs/*_defconfig", the maintainers from the
former block are squashed into the latter.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
The commit (SHA1: 8fc171318e) reintroduced correct values for vendor and
product IDs required for UMS gadget to work properly either at Windows or
Linux.
This data was missing for GONI target, so this commit corrects this mistake.
Signed-off-by: Lukasz Majewski <l.majewski@samsung.com>
Acked-by: Robert Baldyga <r.baldyga@samsung.com>
Acked-by: Minkyu Kang <mk7.kang@samsung.com>
This option is currently not supported, but needs to be, for buildman to
operate as expected.
Reported-by: York Sun <yorksun@freescale.com>
Signed-off-by: Simon Glass <sjg@chromium.org>
snyc with linux v3.15:
commit 1860e379875dfe7271c649058aeddffe5afd9d0d
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date: Sun Jun 8 11:19:54 2014 -0700
Linux 3.15
Signed-off-by: Heiko Schocher <hs@denx.de>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Tom Rini <trini@ti.com>
while playing with the new mtd/ubi/ubifs sync, found some
small updates for it:
- add del_mtd_partition() to include/linux/mtd/mtd
- mtd: add a debug_printf
- remove some not used functions
Signed-off-by: Heiko Schocher <hs@denx.de>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Tom Rini <trini@ti.com>
resync ubi subsystem with linux:
commit 455c6fdbd219161bd09b1165f11699d6d73de11c
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date: Sun Mar 30 20:40:15 2014 -0700
Linux 3.14
A nice side effect of this, is we introduce UBI Fastmap support
to U-Boot.
Signed-off-by: Heiko Schocher <hs@denx.de>
Signed-off-by: Tom Rini <trini@ti.com>
Cc: Marek Vasut <marex@denx.de>
Cc: Sergey Lapin <slapin@ossfans.org>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Joerg Krause <jkrause@posteo.de>
- move linux specific defines from usb and video code
into linux/compat.h
- move common linux specific defines from include/ubi_uboot.h
to linux/compat.h
- add for new mtd/ubi/ubifs sync new needed linux specific
defines to linux/compat.h
Signed-off-by: Heiko Schocher <hs@denx.de>
Cc: Marek Vasut <marex@denx.de>
Cc: Anatolij Gustschin <agust@denx.de>
[trini: Add spin_lock_irqsave/spin_unlock_irqrestore dummies from
usb/lin_gadet_compat.h]
Signed-off-by: Tom Rini <trini@ti.com>
from linux 3.14:
commit 455c6fdbd219161bd09b1165f11699d6d73de11c
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date: Sun Mar 30 20:40:15 2014 -0700
Linux 3.14
Needed for the MTD/UBI/UBIFS resync
Just copied the files from Linux, and added in the c-file
the "#define __UBOOT__" for adding U-Boot special code. In
this case we use this just for adding including U-Boot
headers.
Signed-off-by: Heiko Schocher <hs@denx.de>
Cc: Marek Vasut <marex@denx.de>
Cc: Sergey Lapin <slapin@ossfans.org>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Tom Rini <trini@ti.com>
resync with linux:
commit 455c6fdbd219161bd09b1165f11699d6d73de11c
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date: Sun Mar 30 20:40:15 2014 -0700
Linux 3.14
Needed for the MTD/UBI/UBIFS resync
Just copied the files from Linux, changed the license file header,
and add in the c-file:
+#define __UBOOT__
#include <linux/rbtree_augmented.h>
+#ifndef __UBOOT__
#include <linux/export.h>
+#else
+#include <ubi_uboot.h>
+#endif
so, it compiles for U-Boot.
Signed-off-by: Heiko Schocher <hs@denx.de>
Cc: Marek Vasut <marex@denx.de>
Cc: Sergey Lapin <slapin@ossfans.org>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Tom Rini <trini@ti.com>
The Altera EPCS is SPI flash. We have been using SPI flash driver
to access EPCS for years. The old EPCS driver could be removed.
Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
This patch implements the generic board init as described in
doc/README.generic-board.
Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
Signed-off-by: Scott McNutt <smcnutt@psyent.com>
Reviewed-by: Stefan Roese <sr@denx.de>
When compiling the current code on GCC 4.8.3, the following warnings
appear:
warning: format '%zu' expects argument of type 'size_t', but argument
2 has type 'long unsigned int' [-Wformat=]
There were many mails about such warnings on different architectures.
This patch limits itself to the nios2 architecture.
The problem is that for the size_t (%zu, %zd, ...) arguments of
printf GCC does not verify the type match to size_t type. It verifies
the type match to the compiler-defined __SIZE_TYPE__ type. Thus, if
size_t is defined different from __SIZE_TYPE__ - warnings inevitably
appear.
There is a comment by Thomas Chou to the (rejected) patch:
http://patchwork.ozlabs.org/patch/272102/
which explains that the older GCC toolchains (gcc-3.4.6 and gcc-4.1.2)
expect size_t to be "unsigned long" and the newer expect it to be
"unsigned int". Thus, no matter how we define size_t - either way
warnings appear when using some GCC version.
By rejecting that patch, a choice was made to prefer older GCC versions
and leave the warnings when building with the newer toolchains.
Personally, I disagree with this choice...
In any case, this patch proposes a way to fix the warnings for any GCC
version. Just define size_t using the __SIZE_TYPE__ compiler-defined
type and the type verification will pass.
I tested that this fixes the warning on GCC 4.8.3. I don't have an
older toolchain to test with, but __SIZE_TYPE__ was definitely defined
in GCC 3.4.6, so it should work there too.
Signed-off-by: Vasili Galka <vvv444@gmail.com>
Signed-off-by: Thomas Chou <thomas@wytron.com.tw>
"buildman [options]" is displayed by default.
Append the rest of help messages to parser.usage
instead of replacing it.
Besides, "-b <branch>" is not mandatory since commit fea5858e.
Drop it from the usage.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
It's easier to Cc Simon on patches related to Patman or Buildman.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Cc: Simon Glass <sjg@chromium.org>
"patman [options]" is displayed by default.
Append the rest of help messages to parser.usage
instead of replacing it.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Commit 51148790 added scripts/multiconfig.py written in Python 2
to adjust Kconfig for U-Boot.
It has been hard for Python 3 users because Python 2 and Python 3
are not compatible with each other.
We are not happy about adding a new host tool dependency
(in this case, Python version dependency) for the core build process.
After some discussion, we decided to use only basic tools.
The script may get a bit more unreadable by shell scripting,
but we believe it is worthwhile.
In addition, this commit revives "<board>_config" target that is
equivalent to "<board>_defconfig" for backwards compatibility.
It is annoying to adjust various projects which use U-Boot.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Suggested-by: Igor Grinberg <grinberg@compulab.co.il>
Tested-by: Igor Grinberg <grinberg@compulab.co.il>
Acked-by: Simon Glass <sjg@chromium.org>
Cc: Tom Rini <trini@ti.com>
Cc: Jeroen Hofstee <jeroen@myspectrum.nl>
Cc: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Roger Meier <roger@bufferoverflow.ch>
Reviewed-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Tested-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Cc: Tom Rini <trini@ti.com>
Cc: Wolfgang Denk <wd@denx.de>
Cc: Simon Glass <sjg@chromium.org>
It's not unusual for DHCP servers to take a couple hundred milliseconds
to respond to DHCP discover messages. One possible reason for the delay
can be that the server checks (typically using an ARP request) that the
IP it's about to hand out isn't in use yet. To make matters worse, some
servers may also queue up requests and process them sequentially, which
can cause excessively long delays if clients retry too fast.
Commit f59be6e850 ("net: BOOTP retry timeout improvements") shortened
the retry timeouts significantly, but the BOOTP/DHCP implementation in
U-Boot doesn't handle that well because it will ignore incoming replies
to earlier requests. In one particular setup this increases the time it
takes to obtain a DHCP lease from 630 ms to 8313 ms.
This commit attempts to fix this in two ways. First it increases the
initial retry timeout from 10 ms to 250 ms to give DHCP servers some
more time to respond. At the same time a cache of outstanding DHCP
request IDs is kept so that the implementation will know to continue
transactions even after a retransmission of the DISCOVER message. The
maximum retry timeout is also increased from 1 second to 2 seconds. An
ID cache of size 4 will keep DHCP requests around for 8 seconds (once
the maximum retry timeout has been reached) before dropping them. This
should give servers plenty of time to respond. If it ever turns out
that this isn't enough, the size of the cache can easily be increased.
With this commit the DHCP lease on the above-mentioned setup still takes
longer (1230 ms) than originally, but that's an acceptable compromise to
improve DHCP lease acquisition time for a broader range of setups.
To make it easier to benchmark DHCP in the future, this commit also adds
the time it took to obtain a lease to the final "DHCP client bound to
address x.x.x.x" message.
Tested-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Commit b3dd64f5d5 "bootm: use genimg_get_kernel_addr()" introduced
a bug for booting FIT image. It's because calling fit_parse_config()
twice will give us wrong value in img_addr.
Add a new function genimg_get_kernel_addr_fit() whichl will always
return fit_uname_config and fit_uname_kernel for CONFIG_FIT.
genimg_get_kernel_addr() will ignore those to parameters.
Reported-by: York Sun <yorksun@freescale.com>
Signed-off-by: Bryan Wu <pengw@nvidia.com>
CONFIG_DEFCONFIG_LIST specifies the default defconfig.
It is used by "make savedefconfig" when .config is missing.
But that's it. I could not find other useful cases.
As a side effect, CONFIG_DEFCONFIG_LIST="configs/sandbox_defconfig"
is contained in .config of every target board, which some people
think is odd. So, let's remove it.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Cc: Michal Simek <michal.simek@xilinx.com>
Acked-by: Michal Simek <michal.simek@xilinx.com>
Tom Rini
Stephen Warren
Marek Vasut
Masahiro Yamada
Stefan Roese
Simon Glass
Łukasz Majewski
Heiko Schocher
Thomas Chou
Vasili Galka
Roger Meier
Thierry Reding
Bryan Wu