On Wed, 15 May 2024 12:35:54 +0200, Michael Riesch wrote:
> Habidere,
>
> This series is a follow-up to
> https://lore.barebox.org/barebox/20240412-feature-wolfvision-pf5-v2-0-7e277cc88...@wolfvision.net/
> and fixes a few things. I tried to reply to said thread and announce a
> v3, but I just f
On Wed, 15 May 2024 13:00:37 +0200, Ahmad Fatoum wrote:
> The omission of the 4 was likely a typo. Adding it makes it clearer what
> the function does and serves symmetry with ddr_get_firmware_lpddr4().
>
>
Applied, thanks!
[1/1] ddr: imx: rename ddr_get_firmware_ddr to ddr_get_firmware_ddr4
On Wed, 15 May 2024 13:01:56 +0200, Sascha Hauer wrote:
> The i.MX watchdog has two different configurable behaviours for low
> power modes. The watchdog can either be suspended during low power
> modes or kept running.
> The useful behaviour is normally to disable it during low power modes to
>
On Wed, 15 May 2024 13:19:02 +0200, Ahmad Fatoum wrote:
> The help text erroneously states that CONFIG_FS_PSTORE_RAMOOPS_RO
> disables writing completely, but this is not correct: It merely stops
> barebox from zapping log content previously written by Linux.
>
> The Kconfig help text is correct
On Wed, 15 May 2024 13:20:32 +0200, Ahmad Fatoum wrote:
> If barebox booted a board without state partition on the eMMC it did not set a
> refined compatible at all resulting in a failure to boot into Linux userspace.
>
> Instead just default to the LVDS flavor of the board variant which at leas
On Wed, 15 May 2024 13:22:54 +0200, Ahmad Fatoum wrote:
> We already have a generic place where we check for reserved memory
> matches, so add ramoops there as well.
>
>
Applied, thanks!
[1/1] of: move ramoops device creation into common code
https://git.pengutronix.de/cgit/barebox/comm
On Wed, 15 May 2024 13:27:03 +0200, Ahmad Fatoum wrote:
> This is a port of Linux commit 048a9883267f9b8f8e05dca9e9e8e6f991eea61e:
>
> | Author: Alexey Dobriyan
> | AuthorDate: Sat May 20 21:25:19 2023 +0300
> |
> | include/linux/math.h: fix mult_frac() multiple argument evaluation bug
> |
On Wed, 15 May 2024 15:16:08 +0200, Roland Hieber wrote:
> During the initial installation we can forgo writing an initial state
> variable set to the eMMC if it is zeroed, and if the default values
> priorities are set correctly. Since only the first rootfs partition is
> populated during the in
On Wed, 15 May 2024 17:24:10 +0200, Ahmad Fatoum wrote:
> There's no reason to maintain two allocations per chunk, so just collect
> them both into the same calloc call.
>
> No functional change.
>
>
Applied, thanks!
[1/1] fs: ramfs: allocate once instead of twice per ramfs chunk
https
There's no reason to maintain two allocations per chunk, so just collect
them both into the same calloc call.
No functional change.
Signed-off-by: Ahmad Fatoum
---
I didn't test thoroughly what performance improvement this might bring,
but it looks like a sensible thing to do.
---
fs/ramfs.c |
During the initial installation we can forgo writing an initial state
variable set to the eMMC if it is zeroed, and if the default values
priorities are set correctly. Since only the first rootfs partition is
populated during the install process, it should be preferred over the
second rootfs.
Sign
Hi,
On 15.05.24 07:55, Sascha Hauer wrote:
>
> On Mon, 13 May 2024 16:01:21 +0200, Ahmad Fatoum wrote:
>> A multi_v8 barebox with KASAN enabled is 2051804 bytes even after
>> compression and this breaks linking for me:
>>
>> arch/arm/cpu/common.o: in function `global_variable_offset':
>> arch
This is a port of Linux commit 048a9883267f9b8f8e05dca9e9e8e6f991eea61e:
| Author: Alexey Dobriyan
| AuthorDate: Sat May 20 21:25:19 2023 +0300
|
| include/linux/math.h: fix mult_frac() multiple argument evaluation bug
|
| mult_frac() evaluates _all_ arguments multiple times in the body.
|
|
We already have a generic place where we check for reserved memory
matches, so add ramoops there as well.
Signed-off-by: Ahmad Fatoum
---
drivers/of/base.c | 1 +
fs/pstore/ram.c | 15 ---
2 files changed, 1 insertion(+), 15 deletions(-)
diff --git a/drivers/of/base.c b/drivers/o
From: Soren Andersen
The RAM is to be operated at a slightly higher data rate of 3200 MT/s,
hence the timings have to be adjusted.
Signed-off-by: Soren Andersen
Signed-off-by: Ulrich Ölmann
Signed-off-by: Ahmad Fatoum
---
arch/arm/boards/skov-imx8mp/lpddr4-timing.c | 46 ++---
From: Soren Andersen
If barebox booted a board without state partition on the eMMC it did not set a
refined compatible at all resulting in a failure to boot into Linux userspace.
Instead just default to the LVDS flavor of the board variant which at least
leads to a working Linux userspace.
Sign
The help text erroneously states that CONFIG_FS_PSTORE_RAMOOPS_RO
disables writing completely, but this is not correct: It merely stops
barebox from zapping log content previously written by Linux.
The Kconfig help text is correct, so adjust the documentation to reflect
that and while at it, add s
The i.MX watchdog has two different configurable behaviours for low
power modes. The watchdog can either be suspended during low power
modes or kept running.
The useful behaviour is normally to disable it during low power modes to
be able to suspend the system without having the watchdog resetting
The omission of the 4 was likely a typo. Adding it makes it clearer what
the function does and serves symmetry with ddr_get_firmware_lpddr4().
Signed-off-by: Ahmad Fatoum
---
drivers/ddr/imx/ddrphy_train.c | 2 +-
include/soc/imx/ddr.h | 4 ++--
2 files changed, 3 insertions(+), 3 delet
are the fixes. Do you think they can be applied directly
to master?
Looking forward to your comments!
Signed-off-by: Michael Riesch
---
Changes in v2:
- replace basprintf() with xstrdup() to avoid compiler warning
- Link to v1:
https://lore.kernel.org/r/20240515-b4-pf5-fixup-v1-0-a58c8ec94
Since hardware ID detection has been moved to the WolfVision board code
library, move the Kconfig dependencies as well.
Signed-off-by: Michael Riesch
---
arch/arm/mach-rockchip/Kconfig | 2 --
common/boards/Kconfig | 2 ++
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/a
Add pr_fmt() message prefix to WolfVision PF5 board code.
Signed-off-by: Michael Riesch
---
arch/arm/boards/wolfvision-pf5/board.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/arch/arm/boards/wolfvision-pf5/board.c
b/arch/arm/boards/wolfvision-pf5/board.c
index 797f51bc2e..5a2f4201ba
If the char **overlays parameter to wolfvision_rk3568_detect_hw is NULL,
the overlay file names are not collected.
If overlays points to a NULL pointer, it is initialized properly with
an empty string. This is convenient as the call to
globalvar_set("of.overlay.filepattern", my_resulting_strin
Hi Ahmad,
On 5/15/24 11:14, Ahmad Fatoum wrote:
> On 15.05.24 10:32, Michael Riesch wrote:
>> Hi all,
>>
>> On 5/15/24 10:00, Michael Riesch wrote:
>>> If the char **overlays parameter to wolfvision_rk3568_detect_hw is NULL,
>>> the overlay file names are not collected.
>>>
>>> If overlays points
partitions: add prompt and help text for CONFIG_PARTITION_MANIPULATION
So far, CONFIG_PARTITION_MANIPULATION was only selected by
CONFIG_CMD_PARTED and influenced repartitioning support.
Now that it's used to guard the feature of reparsing partition tables,
it makes sense to give it a help text a
On 15.05.24 10:32, Michael Riesch wrote:
> Hi all,
>
> On 5/15/24 10:00, Michael Riesch wrote:
>> If the char **overlays parameter to wolfvision_rk3568_detect_hw is NULL,
>> the overlay file names are not collected.
>>
>> If overlays points to a NULL pointer, it is initialized properly with
>> an
Hi all,
On 5/15/24 10:00, Michael Riesch wrote:
> If the char **overlays parameter to wolfvision_rk3568_detect_hw is NULL,
> the overlay file names are not collected.
>
> If overlays points to a NULL pointer, it is initialized properly with
> an empty string. This is convenient as the call to
>
reparse_partition_table() will delete a device's existing partition
cdevs and allocate new ones according to the new partition table.
Holding a reference to a cdev prevents it from bring removed to avoid
dangling pointers and use-after-frees.
Unfortunately, not all users call cdev_open on the cde
cdev_operations::close can fail and thus cdev_close should pass along
the error code to the users instead of discarding it.
Do that, so users like devfs close() can start propagating this error.
Signed-off-by: Ahmad Fatoum
---
fs/devfs-core.c | 11 ---
include/driver.h | 2 +-
2 files
barebox v2024.03.0 moves the reference count maintenance into
cdev_open/cdev_close, but missed to switch devfs_open and devfs_close to
actually use cdev_open/cdev_close. The result was that the cdev could be
deleted, e.g. due to partition reparsing, and further use of the file
descriptor would resu
If the char **overlays parameter to wolfvision_rk3568_detect_hw is NULL,
the overlay file names are not collected.
If overlays points to a NULL pointer, it is initialized properly with
an empty string. This is convenient as the call to
globalvar_set("of.overlay.filepattern", my_resulting_strin
Since hardware ID detection has been moved to the WolfVision board code
library, move the Kconfig dependencies as well.
Signed-off-by: Michael Riesch
---
arch/arm/mach-rockchip/Kconfig | 2 --
common/boards/Kconfig | 2 ++
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/a
Add pr_fmt() message prefix to WolfVision PF5 board code.
Signed-off-by: Michael Riesch
---
arch/arm/boards/wolfvision-pf5/board.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/arch/arm/boards/wolfvision-pf5/board.c
b/arch/arm/boards/wolfvision-pf5/board.c
index 797f51bc2e..5a2f4201ba
deletions(-)
---
base-commit: 593248cde35ddedcb27c0791c621e6a4403d7068
change-id: 20240515-b4-pf5-fixup-1c0f8b592313
Best regards,
--
Michael Riesch
34 matches
Mail list logo