GitLab

According to fsync specification [1] some special files (e.g., a pipe, FIFO,
or socket) don't support synchronization and return either EROFS or EINVAL.

On the linux side the sys_fsync -> do_fsync() checks if the requested file
has f_op->fsync defined. If not it returns EINVAL [2].

This commit prevents writing error messages for files (devices), which
do not support fsync().

[1] - http://man7.org/linux/man-pages/man2/fsync.2.html
[2] - http://elixir.free-electrons.com/linux/v4.13-rc6/source/fs/sync.c#L183

Signed-off-by: Lukasz Majewski <lukma@denx.de>
Acked-by: Michael Heimpold <mhei@heimpold.de>

We have a board with several revisions. The older ones use a nor flash
with 64k erase size, while the newer have a flash with 4k sectors. The
environment size is 8k.

Currently, we have to put a column containing 0x10000 (64k) in
fw_env.config in order for it to work on the older boards. But that
ends up wasting quite a lot of time on the newer boards that could
just erase the 8k occupied by the environment - strace says the 64k
erase takes 0.405 seconds. With this patch, as expected, that's about
an 8-fold better, at 0.043 seconds.

Having different fw_env.config files for the different revisions is
highly impractical, and the correct information is already available
right at our fingertips. So use the erasesize returned by the
MEMGETINFO ioctl when the fourth and fifth columns (sector size and
#sectors, respectively) are absent or contain 0, a case where the
logic previously used to use the environment size as erase size (and
consequently computed ENVSECTORS(dev) as 1).

As I'm only testing this on a NOR flash, I'm only changing the logic
for that case, though I think it should be possible for the other
types as well.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>

Closing a file descriptor does not guarantee that the data has been
successfully saved to disk, as the kernel might defer the write.
Signed-off-by: Michael Heimpold <mhei@heimpold.de>

Initially to detect the flash, DDR mode is not used so the TDH bits
need to be cleared. In case of booting from QSPI, the BootROM might
have left the TDH bits set, so we need to clear them actively.
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>

The write data size can be overwritten by writing to the IDATSZ
field of IPCR register. Since the driver always updates the IDATSZ
in page program operation. Set the LUT data size to 0 to align
the codes with iMX.
Signed-off-by: Ye Li <ye.li@nxp.com>

On latest iMX platforms like iMX7D/iMX6UL/iMX8MQ, the QSPI controller
is updated to have TDH field in FLSHCR register. According to reference
manual, this TDH must be set to 1 when DDR_EN is set. Otherwise, the TX
DDR delay logic won't be enabled.

Another issue in DDR mode is the MCR register will be overwritten in
every read/write/erase operation. This causes DDR_EN been cleared while
TDH=1, then no clk2x output for TX data shift and all operations will
fail.
Signed-off-by: Ye Li <ye.li@nxp.com>

Use readl_poll_timeout instead of explicit calculation
Signed-off-by: Thomas Schaefer <thomas.schaefer@kontron.com>
Reviewed-by: Fabio Estevam <festevam@gmail.com>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>

During QSPI reads, current is_controller_busy function sporadically
fails with -ETIMEDOUT due to fixed number of 5 test loops. Using
timer functions to wait 1000 us instead will fix this.
Signed-off-by: Thomas Schaefer <thomas.schaefer@kontron.com>
Reviewed-by: Fabio Estevam <festevam@gmail.com>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>

In

	void *rx_addr = NULL;
	rx_add = A;

the first assignment has no effect. Remove it.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Jagan Teki <jagan@openedev.com>

Some SoCs have different endianness of QSPI IP if compared
to endianness of core. The function is_controller_busy()
checks if the QSPI controller is busy or not, considering
the endianness of the QSPI IP.
Signed-off-by: Rajat Srivastava <rajat.srivastava@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>

Part of upstream commit 48263504c8d501678acaa90c075f3f7cda17c316.

The QSPI module on i.MX7D is modified from i.MX6SX. The module used on
i.MX6UL/6ULL is reused from i.MX7D. They share same tx buffer size.

The endianness is not set at qspi driver initialization. So if we don't
boot from QSPI, we will get wrong endianness when accessing from AHB
address directly.

Add the compatible entry for 6ul/7d.
Signed-off-by: Peng Fan <peng.fan@nxp.com>

In some of the QSPI controller version, there must be atleast
128bit data available in TX FIFO for any pop operation otherwise
error bit will be set. The code will not make any behavior change
for previous controller as the transfer data size in ipcr register
is still the same.

Patch is tested on LS1046A which do not require 16 bytes aligned and
LS1088A which require 16 bytes aligned data in TX FIFO
Signed-off-by: Suresh Gupta <suresh.gupta@nxp.com>
Signed-off-by: Anupam Kumar <anupam.kumar_1@nxp.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>

It is recommended to check either controller is free to take
new spi action. The IP_ACC and AHB_ACC bits indicates that
the controller is busy in IP or AHB mode respectively.
And the BUSY bit indicates that controller is currently
busy handling a transaction to an external flash device
Signed-off-by: Suresh Gupta <suresh.gupta@nxp.com>
Reviewed-by: Jagan Teki <jagan@openedev.com>

Pet the watchdog once upon each command call (qspi_xfer) and during
each loop iteration in several commands.

This fixes a watchdog reset especially during erase command.
Signed-off-by: Alexander Stein <alexander.stein@systec-electronic.com>
Reviewed-by: York Sun <york.sun@nxp.com>

SOC’s like LS1012A has only one chip select signal for QSPI flash.
Avoid scanning other flash.
Signed-off-by: Suresh Gupta <suresh.gupta@nxp.com>
Reviewed-by: York Sun <york.sun@nxp.com>

PXE boot is broken with GCC 7.1 due option '-fstore-merging' enabled
by default for '-O2':

BOOTP broadcast 1
data abort
pc : [<8ff8bb30>]          lr : [<00004f1f>]
reloc pc : [<17832b30>]    lr : [<878abf1f>]
sp : 8f558bc0  ip : 00000000     fp : 8ffef5a4
r10: 8ffed248  r9 : 8f558ee0     r8 : 8ffef594
r7 : 0000000e  r6 : 8ffed700     r5 : 00000000  r4 : 8ffed74e
r3 : 00060101  r2 : 8ffed230     r1 : 8ffed706  r0 : 00000ddd
Flags: nzcv  IRQs off  FIQs off  Mode SVC_32
Resetting CPU ...

Core reason is usage of structures for network headers without packed
attribute.
Reviewed-by: Yauheni Kaliuta <yauheni.kaliuta@redhat.com>
Signed-off-by: Denis Pynkin <denis.pynkin@collabora.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

By checking ubifs source code, s_instances parameter is not
used anymore. So, set this parameter and the associated source
code under __UBOOT__ compilation.

This also fixes a real bug, as with multiple mount operations,
the update of the fs_supers linked list can cause memory corruption.

Upstream commit: 5a08cfee3967
Signed-off-by: Christophe Kerello <christophe.kerello@st.com>
Signed-off-by: Patrice Chotard <patrice.chotard@st.com>

Clear powerdown enable bit for WDOG3 on i.MX6ULL to avoid unwanted
kernel reboots.
Suggested-by: Ye Li <ye.li@nxp.com>
Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Stefano Babic <sbabic@denx.de>

i.MX6ULL has the same WDOG3 base address as i.MX6UL, so take this
into account in the base address definition.
Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Stefano Babic <sbabic@denx.de>

Make it possible to use first, second and last page for bad block markers

See merge request !1

Some ESMT NANDs have been shipped by the manufacturer with bad block
markers in first or last page, instead of first or second page.
Also the datasheets for Cypress/Spansion/AMD NANDs claim that the
first, second and last page needs to be checked.
Therefore we need to scan all three of those pages to make sure to
detect the markers correctly for these manufacturers.
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>

Add the manufacturer id of ESMT NANDs to apply manufacturer specific
configuration.
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>

Currently supported bad block marker positions within the block are:
* in first page only
* in last page only
* in first or second page

Some ESMT NANDs have been shipped by the manufacturer with bad block
markers in first or last page.
Also the datasheets for Cypress/Spansion/AMD NANDs claim that the
first, second and last page needs to be checked.
Therefore we make it possible to set NAND_BBT_SCAN2NDPAGE and
NAND_BBT_SCANLASTPAGE at the same time to scan/set all three pages.
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>

fix curl call

For proper authentication users must check the IVT before authenticating the
kernel image. If the IVT DCD pointer is not 0, display an error as shown below
and return a boot failure.

	Authenticate image from DDR location 0x12000000...
        Error: DCD pointer must be 0

        Authenticate zImage Fail, Please check

Backported to branch imx_v2015.04_4.1.15_1.0.0_ga.
Signed-off-by: Utkarsh Gupta <utkarsh.gupta@nxp.com>
Signed-off-by: Breno Lima <breno.lima@nxp.com>

For proper authentication users must check if the CSF is valid. This patch
calls the csf_is_valid function prior to authenticating the image to parse
the CSF and returns a failure if an invalid command is found as shown below.

	Authenticate image from DDR location 0x12000000...
	Check CSF for Write Data command before authenticating image

	Error - WRITE Data command found

	Authenticate zImage Fail, Please check

Backported to branch imx_v2017.03_4.9.11_1.0.0_ga.
Signed-off-by: Utkarsh Gupta <utkarsh.gupta@nxp.com>

Only part of the data was written to flash due to the lacking of
program load (0x02) instruction

set variable ethaddr only for first eth device.
-> solves wrong address enumeration