Reserve memory defined through "reg" is unconditionally set as reserve
memory, so these reserve memory regions can also be successfully set in
the crash kernel. However, if they are also used as CMA memory and
their starting addresses are not within the accessible range of the
crash kernel, it will
Atomic operations should be employed within rwlock, hence GFP_ATOMIC
is utilized instead of GFP_KERNEL. This ensures the avoidance of the
subsequent call trace and prevents potential misinterpretation of NPU
performance degradation during testing.
BUG: sleeping function called from invalid context
commit 328efccc7847 ("soc: imx8m: Enable OCOTP clock for imx8mm
before reading registers") has re-enabled the clock, so reading
the registers needs to be done before clk_disable_unprepare(),
otherwise the kernel will hang.
Fixes a3985483e7f1 ("LF-9486 soc: imx8mp: support 128 bits UID")
Signed-off
A bug was introduced by commit 20932f1f83d3 ("fec_ecat: add fec native
driver for raw packet proto"). When both fd and fast_raw_socket_fd are
-1, the fast_raw_socket_dev would be null, then it will perform a null
pointer operation and causes the following oops. The fast_raw_socket_dev
is only set t
gpu: imx: dpu: Use raw_spin_lock instead of mutex_lock
Xiaolei Wang via lists.yoctoproject.org (1):
imx: dpu-blit: Remove GFP_DMA32
arch/arm/boot/dts/nxp/imx/imx6q-sabresd.dts | 4 +
arch/arm/mach-imx/Kconfig| 1 +
arch/arm/mach-imx/Makefile | 2 +-
arch
For imx6 bsp, the SDK actually integrates imx6 config
and imx7 config, so lpddr3_freq_imx is needed when we
use imx6, otherwise it will cause a series of problems,
such as blurred graphics displayed by the display system,
etc., so in order to avoid turning on the CONFIG_SOC_IMX7D and
CONFIG_SOC_IMX
Changeset application failure should be destroyed
unreferenced object 0x0008f2656480 (size 128):
comm "kworker/2:1", pid 83, jiffies 4294893570 (age 70.724s)
hex dump (first 32 bytes):
c0 b6 db 11 00 80 ff ff c0 b6 db 11 00 80 ff ff
02 00 00 00 00 00 00 00 28 9e 4
Changeset application failure should be destroyed
unreferenced object 0x0008f249a980 (size 128):
comm "kworker/2:1", pid 83, jiffies 4294893542 (age 70.832s) hex dump
(first 32 bytes):
e0 b6 db 11 00 80 ff ff e0 b6 db 11 00 80 ff ff
05 00 00 00 00 00 00 00 20 de 44 ff 08
The irqsteer's runtime PM has to be enabled otherwise the probe of
some devices (such as dpu) will fail because of the failure of
irq_chip_pm_get().
But we can't allow the irqsteer to really enter the suspend state
because it will break the kdump with a call trace like below.
Internal error: sy
CPU: 0 PID: 0 Comm: swapper/0 Tainted: GW
5.15.78-rt48-yocto-preempt-rt #1
Hardware name: Freescale i.MX7 Dual (Device Tree)
[<80110f4c>] (unwind_backtrace) from [<8010aef8>] (show_stack+0x18/0x1c)
[<8010aef8>] (show_stack) from [<80be4998>] (dump_stack_lvl+0x40/0x4c)
[<80be4998>]
Refer to commit 323751dadfbe("MLK-21958-13: drm/bridge: adv7511: Limit
supported clocks")
Add Limit supported clocks for adv7533, Some modes are not working with this
converter.
This will cause hdmi link down, so add this workround patch to solve this
problem.
Signed-off-by: Xiaolei Wang
Signe
When we don`t set "video=mxcfb0:dev=hdmi" in bootargs,
but connect hdmi to imx6q, it will trigger a periodically
hdmi interrupt in a few seconds. These repeated interruptions
will read edid and alloc modedb, so we need to release the
modedb before the next allocation.
Fix following memory leak.
Enable pcie for imx6q-sabresd.
Signed-off-by: Xiaolei Wang
Signed-off-by: Bruce Ashfield
---
arch/arm/boot/dts/nxp/imx/imx6q-sabresd.dts | 4
1 file changed, 4 insertions(+)
diff --git a/arch/arm/boot/dts/nxp/imx/imx6q-sabresd.dts
b/arch/arm/boot/dts/nxp/imx/imx6q-sabresd.dts
index ca8c7
When mxc_hdmi_disp_init returns failure, hdmi->fbi->modelist needs
to be released in mxc_hdmi_disp_deinit(), this patch solves the
following memory leaks:
unreferenced object 0xc52a45c0 (size 64):
comm "swapper/0", pid 1, jiffies 4294937844 (age 159.630s)
hex dump (first 32 bytes):
80 45 2
The data structure edid_fbi needs to be initialized, otherwise
edid_fbi->monspecs.modedb and fbi->monspecs.modedb_len are not
initial values, which will cause illegal free to be detected
by KASAN, and the following panic will appear:
Unable to handle kernel paging request at virtual address 80f5e0
Use raw_spin_lock instead of spin_lock,
this patch fixes the following warning:
Hardware name: Freescale i.MX7 Dual (Device Tree)
unwind_backtrace from show_stack+0x18/0x1c
show_stack from dump_stack_lvl+0x40/0x4c
dump_stack_lvl from __might_resched+0x140/0x1b4
__might_resched from rt_spin_loc
Use raw_spin_lock instead of spin_lock,
this patch fixes the following warning:
BUG: sleeping function called from invalid context at
kernel/locking/spinlock_rt.c:46
in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 506, name: sh
preempt_count: 2, expected: 0
RCU nest depth: 0, expected: 0
From: "Xiaolei Wang via lists.yoctoproject.org"
The DMA32 zone on imx8qm is empty, and GFP_DMA32 and
GFP_DMA are a bad combination.
Signed-off-by: Xiaolei Wang
Signed-off-by: Bruce Ashfield
---
drivers/gpu/imx/dpu-blit/dpu-blit.c | 2 +-
1 file changed, 1 insertion(+), 1 deletio
For Atomic Mode Setting, it is safer to use raw spin lock.
Mutex may cause scheduling, which avoids the following warning.
BUG: sleeping function called from invalid context at kernel/locking/mutex.c:283
in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 191, name: kworker/5:13
preempt_count
new scc and cfg files for BSP nxp-imx6 on branch yocto-6.6.
Signed-off-by: Xiaolei Wang
---
bsp/nxp-imx6/nxp-imx6-preempt-rt.scc | 7 +
bsp/nxp-imx6/nxp-imx6-standard.scc | 7 +
bsp/nxp-imx6/nxp-imx6.cfg| 226 +++
bsp/nxp-imx6/nxp-imx6.scc|
imx6 memory range 0x8000 - 0xC000, arm32
itself supports 2G/2G partitioning. If it is set
to 3G/1G partitioning, part of the memory will be
allocated to highmem. imx6 needs to save 320M cma
memory. When cma is active, if pfn is not in a zone,
it will fail. Since cma has reserved a range, wh
From: Xiaolei wang
When probe fails, disable runtime, commit 6eb1453fde64
("LF-106: ASoC: fsl_ssi: request BUS_FREQ_AUDIO") introduces
a problem. When PROVE_LOCKING is turned on, the system will
randomly freeze at startup. This is related to AUDIO BUSFREQ.
Since IMX_SDMA is compiled as 'M', fsl_s
101 - 122 of 122 matches
Mail list logo
