| Age | Commit message (Collapse) | Author |
|---|
|
Surface 2 is a Surface-series Windows RT hybrid tablet computer created by
Microsoft. Surface 2 uses a 1.7 GHz quad-core Nvidia Tegra 4 chipset with
2 GB of RAM, features 10.6 inch FullHD ClearType HD screen with 16:9 aspect
ratio and 32/64 GB of internal memory that can be supplemented with a
microSDXC card giving up to 64 GB of additional storage.
Signed-off-by: Jonas Schwöbel <[email protected]>
Signed-off-by: Svyatoslav Ryhel <[email protected]>
|
|
Neha Malcom Francis <[email protected]> says:
Currently, the inline ECC implementation enables inline ECC across the
entire DDR space. However this is not always required and a more common
ask is to have only a portion of the DDR protected as enabling ECC
impacts read/write performance metrics.
This series aims to modify the logic to firstly support partial inline
ECC in its' most basic form which works for single controllers. Then it
introduces an algorithm to support multi DDR controllers where
interleaving plays a role. Since interleaving is handled by the MSMC, it
only makes sense to have the MSMC decide the inline ECC ranges for each
DDR.
This series also introduces support for multiple partial regions of inline
ECC however due to complexity only support for single DDR is present now.
WIP: A commandline test case patch for verifying the correct behaviour
of inline ECC including partial case. Was targeted for v2 however a little
tricky to make it a general test case especially for multi-DDR cases, so
have not combined it in this series for now.
Testing:
- Memtester runs for J721S2 and J784S4 platforms with and without ECC
enablement runs fine.
- Along with patches that add support for the commandline test (see WIP
note above) J784S4 shows expected behavior for three sets of partial
inline ECC regions (non-overlapping, and after modifying J784S4 to
have single DDR instead of multi-DDR):
https://gist.github.com/nehamalcom/bde7e14e96485e4a188c3af3af6d75d6
Link: https://lore.kernel.org/r/[email protected]
|
|
K3 Inline ECC mechanism can support up to 3 regions of inline ECC, add
this support for single controller.
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
The existing approach does not account for interleaving in the DDRs when
setting up regions. There is support for MSMC to calculate the regions
for each DDR, so modify k3_ddrss_probe to set the regions accordingly
for multi-DDR systems.
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
Add support for calculation of the protected regions for each DDR in
multi-DDR systems. Since MSMC is the parent node of the individual DDRs
as well as responsible for their interleaving, it only makes sense for
MSMC to contain the logic for dividing the regions.
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
In K3 multi-DDR systems, the MSMC is responsible for the interleave
mechanism across all the DDR controllers. Add support for MSMC to obtain
the number of controllers it's responsible for using the DT.
Reviewed-by: Udit Kumar <[email protected]>
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
As we increase the functionalities that the K3 DDRSS sub-system support,
it is becoming more evident that the same logic cannot apply to both
single as well as multiple DDR controller devices. Add
CONFIG_K3_MULTI_DDR to be used to differentiate between the two.
Reviewed-by: Udit Kumar <[email protected]>
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
Instead of defaulting to choosing the entire DDR region when enabling
inline ECC, allow picking of a range within the DDR space using DT to
enable.
It expects such a node within the memory node, in the absence of which
we resort to enabling inline ECC for the entire DDR region:
inline_ecc: protected@9e780000 {
device_type = "ecc";
reg = <0x9e780000 0x0080000>;
bootph-all;
};
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
The reserved space needed for storing the parity remains the same no
matter the size of the region that is being protected. Add this as a
comment for better code understanding.
Reviewed-by: Udit Kumar <[email protected]>
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
To prevent checkpatch warning once we start using this macro more
frequently, shorten the length of it. While at it, also move the
structure k3_ddrss_ecc_region above k3_msmc so that future patches can
have it as a member of k3_msmc.
Reviewed-by: Udit Kumar <[email protected]>
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
Let ecc_regions[x].start reflect the start of the ECC region in terms of
DDR addressing rather than system addressing. This will make it easier
to extend the usage of the same ecc_regions structure for multi-DDR
systems as well.
Reviewed-by: Udit Kumar <[email protected]>
Signed-off-by: Neha Malcom Francis <[email protected]>
|
|
Andrew Goodbody <[email protected]> says:
Smatch reported issues including a derference of a pointer before its
NULL check and the use of an uninitialised variable.
Link: https://lore.kernel.org/r/[email protected]
|
|
In k3_nav_ringacc_probe_dt there can be no error code returned from
dev_read_u32_default so ret is not assigned to and should not be used.
Remove the use of ret from the dev_err call as it is unitialised.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
Move the first dereference of ring to after the NULL check has occurred.
This will prevent any possible dereference of NULL.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
Andrew Goodbody <[email protected]> says:
Smatch reported two issues, firstly attempting to compare a u8 to a 16
bit macro and secondly a potentially uninitialised variable.
Link: https://lore.kernel.org/r/[email protected]
|
|
In k3_r5f_split_reset and k3_r5f_unprepare ret may not have been
assigned to before the code reaches the return ret at the function exit.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
In the struct ti_sci_proc the fields proc_id and host_id are declared as
u8 so cannot be set to nor compared with a macro defined with a value
using 16 bits. Change the macro to only use 8 bits to make the code work
as expected.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
In order to compile code that uses <asm/atomic.h> on sandbox, we must
provide this header. RISC-V shows us today how to do so with the generic
header implementation, so copy that.
Signed-off-by: Tom Rini <[email protected]>
|
|
Normally, local_save_flags is used as part of the local_irq_* macros, so
remove that as it's unused. Make local_irq_save do something to the
passed variable so that it won't trigger unused variable warnings later.
Signed-off-by: Tom Rini <[email protected]>
|
|
In rt5677_bic_or the call to rt5677_i2c_read returns an int so old
should also be an int to receive that value and then be able to test it
for being negative which would indicate an error.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
We are trying to enable -Wmissing-prototypes and this functiion is only
used locally. Mark it as static.
Signed-off-by: Ilias Apalodimas <[email protected]>
|
|
The function is only used locally so declare it as static.
Signed-off-by: Ilias Apalodimas <[email protected]>
|
|
In s5pxx18_pinctrl_set_state testing count to be negative will always
fail as count is unsigned despite receiving the return value of a
function that returns an int. Change count and idx to be of type int to
allow the test to work as expected and remove the need for any implicit
casts. Also change pin to be u32 which is what all called functions
expect.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
This series from Ilias Apalodimas <[email protected]> fixes a
number of issues with running the EFI capsule tests in CI.
Link: https://lore.kernel.org/r/[email protected]
|
|
Capsule updates tests have been skipped since
commit 659f97eb1fc3 ("scripts/Makefile.lib: EFI: Use capsule CRT instead of ESL file")
Remove that check since it's not needed anymore and re-enable the tests.
Fixes: 659f97eb1fc3 ("scripts/Makefile.lib: EFI: Use capsule CRT instead of ESL file")
Signed-off-by: Ilias Apalodimas <[email protected]>
|
|
efi_capsule_data() is called in each of the EFI tests to create and
setup the files we need. However, it also recreates the spi.bin file
that holds the SPI flash contents we rely on for the test validation.
This leads to weird errors since reading from the flash returns 0,
instead of the expected value if the file has been recreated.
Always restart our sandbox instance if the files are recreated.
Signed-off-by: Ilias Apalodimas <[email protected]>
|
|
The current code writes values to a flash offset defined by a function
argument. However, when reading it back we always read from a static
offset. Adjust the reads to use the correct offset.
Signed-off-by: Ilias Apalodimas <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
|
|
Some capsule tests are changing the sandbox DT to test various features,
e.g authenticated capsule updates, versioning support etc. However, no one
restores the original DT and the CI pops errors looking like
/u-boot
Bloblist at 100 not found (err=-2)
Failed to find FDT file '/tmp/sandbox/persistent-data/scratch/EFI/CapsuleTestData/test_ver.dtb'
initcall_run_f(): initcall fdtdec_setup() failed
if sandbox is restarted.
So let's restore the proper DT after done with the capsule testing.
Signed-off-by: Ilias Apalodimas <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
|
|
Jim Liu <[email protected]> says:
modify npcm7xx/8xx feature and bug fixed
Link: https://lore.kernel.org/r/[email protected]
|
|
Set the uart clock frequency according to dts by default.
If CONFIG_SYS_SKIP_UART_INIT is not enabled, no need to
do board_set_console to change the console bootarg.
Signed-off-by: Jim Liu <[email protected]>
|
|
If there is an unhandled KCS/BPC pending interrupt after reboot,
the KCS/BPC Linux driver may trigger interrupts immediately upon
registering the irq. However, since the driver is not yet initialized
to handle them, this can lead to unexpected behavior.
To prevent this, disable KCS/BPC interrupts in u-boot to avoid pending
interrupts from being raised before the Linux driver is fully initialized.
Signed-off-by: Stanley Chu <[email protected]>
Signed-off-by: Jim Liu <[email protected]>
|
|
Enabling an eSPI channel(e.g. Peripheral Channel) during
an eSPI transaction might cause the BMC eSPI module to
transition to a wrong state and therefore respond with
FATAL_ERROR on incoming transaction.
Add workaround to avoid the module getting into the wrong
state.
Signed-off-by: Stanley Chu <[email protected]>
Signed-off-by: Jim Liu <[email protected]>
|
|
npcm_timer driver is changed to use SECCNT counter.
Signed-off-by: Jim Liu <[email protected]>
|
|
When doing a dm_i2c_read followed by a dm_i2c_write, the subsequent
transaction may get npcm_i2c_check_sda error because the module is
still busy in STOP condition in previous dm_i2c_read.
Always check and wait for module to be out of busy before starting
an i2c transaction.
Signed-off-by: Stanley Chu <[email protected]>
Signed-off-by: Jim Liu <[email protected]>
|
|
Add pinmux for the VCD input to use the HSYNC signal.
Signed-off-by: Stanley Chu <[email protected]>
Signed-off-by: Jim Liu <[email protected]>
|
|
Add pinmux to select the HSYNC signal as the VCD input.
Signed-off-by: Stanley Chu <[email protected]>
Signed-off-by: Jim Liu <[email protected]>
|
|
Pull request net-20250818.
This is mostly code cleanup and fixes, mainly for issues reported by the
Smatch tool, plus two small features for NET_LWIP as well as support for
the BCM54612E phy.
net-common:
- Fix a bunch of issues reported by Smatch
- Introduce CONFIG_DNS
- Add support for BCM54612E phy
net-legacy:
- Add missing SPDX-License-Identifier for files originating from LiMon
net-lwip:
- ping: initialize net_try_count to 1
- sntp: remove redundant sys_check_timeouts()
- tftp: resend initial request
- Add Kconfig option to show ICMP unreachable errors
|
|
In cdns_torrent_phy_probe the test for too many lanes configured does
not set an error code before taking the error path. This could lead to a
silent failure if the calling code does not detect the error. Add the
code to return -EINVAL in this case.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
In cdns_sierra_pll_bind_of_clocks the variable 'i' is declared but never
assigned to before its value is used in a dev_err. Replace clk_names[i]
by the name passed to device_bind(), i.e., "pll_mux_clk". With that, the
clk_names[] array is unused and can therefore be removed.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
[jf: update description]
Signed-off-by: Jerome Forissier <[email protected]>
|
|
In ks8851_mll_detect_chip the if..else code detects the case of (val &
0xfff0) != CIDER_ID and returns if found. So testing for this again will
always fail and the code is unreachable. Just remove the test and code
block.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
The legacy network stack sets net_try_count to 1 at the beginning of the
net_loop() function. This is required for net_start_again() to work
properly. Therefore, set the variable accordingly in the do_ping()
function when NET_LWIP=y. This fixes an issue where a ping to an
unreachable destination would run twice on the same network device. For
example with qemu_arm64_lwip_defconfig:
=> dhcp
DHCP client bound to address 10.0.2.15 (3 ms)
=> ping 10.0.0.1
Using virtio-net#32 device
ping failed; host 10.0.0.1 is not alive
Using virtio-net#32 device
ping failed; host 10.0.0.1 is not alive
=> QEMU: Terminated
Signed-off-by: Jerome Forissier <[email protected]>
|
|
In ca_phy_probe when checking for an external phy it uses a field from
the internal phy due to what is assumed to be a copy/paste typo. Make
the obvious fix to use the field from the external phy.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
Now that sys_check_timeouts() is called in net_lwip_rx(), there is no
need to call it from the SNTP receive loop. Remove the redundant call.
Signed-off-by: Jerome Forissier <[email protected]>
|
|
The TFTP implementation does not resend the initial request if there is
no response from the server. Since TFTP is based on UDP, there should be
a mechanism to deal with unreliable transmissions at this point, similar
to what we have for data packets. Therefore, introduce request
retransmission.
Signed-off-by: Jerome Forissier <[email protected]>
CC: Venkatesh Abbarapu <[email protected]>
CC: Michal Simek <[email protected]>
CC: Ilias Apalodimas <[email protected]>
|
|
Introduce the DNS Kconfig symbol so that various network commands may
use host names without the dns command (CMD_DNS) being selected.
Signed-off-by: Jerome Forissier <[email protected]>
CC: E Shattow <[email protected]>
|
|
In rvu_af_init if the code fails to allocate memory for nix_af it will
take the error path with nix_af == NULL which will dereference nix_af.
Add the appropriate NULL check.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
In nix_lf_setup there is a default value assigned to err in case an
error is detected. However this default value will be overwritten in the
for loop so that later code does not return an error code from the
function. Add a new assignment to restore err to the default error code.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
In mvpp2_probe the code attempts to get a value for "gop-port-id" and
assigns it to port->gop_id but it then tests port->id for being equal to
-1. That is an impossible test as port->id is a field of type u8 so
cannot be negative. Change the test to port->gop_id.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
In phy_info_parse all uses of the variable phyaddr are as an int so
declaring as u32 is not useful and prevents the test for an error return
from fdtdec_get_int ever detecting an error. Change phyaddr to be an
int.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
|
Instead of returning -1 on a failed alloc, return -ENOMEM.
This issue was found by Smatch.
Signed-off-by: Andrew Goodbody <[email protected]>
|
