Kernel/aarch64: Get RPi::InterruptController address and RPi::Timer address + IRQ number from the devicetree #25500
Conversation
github-actions
bot
added
the
👀 pr-needs-review
spholz
force-pushed
the
devicetree-mmio
branch
2 times, most recently
from
902f738
to
751568f
Compare
| } | ||
|
|
||
| private: | ||
| Vector<u8, 4 * sizeof(u32)> m_raw; |
There was a problem hiding this comment.
FixedArray should fit a bit better here
There was a problem hiding this comment.
I use Vectors because FixedArrays don't have an inline capacity.
There was a problem hiding this comment.
Then maybe ByteBuffer?
Vector feels wrong, but works I guess
There was a problem hiding this comment.
Then I would have to use Detail::ByteBuffer. ::ByteBuffer is just an alias for Detail::ByteBuffer<32>.
An inline capacity of 32 seems wasteful.
There was a problem hiding this comment.
I could move ByteBuffer out of Detail and change it to have a default value (Ali suggested that in the PR that introduced the current ByteBuffer as well: #7151 (comment))
I'm not sure if this would cause any problems. But we would have to replace all current occurrences of ByteBuffer with ByteBuffer<>, which probably would be annoying for cherry-picking ladybird commits I think (@nico)
| } | ||
|
|
||
| private: | ||
| Vector<u8, 2 * sizeof(u32)> m_raw; |
spholz
force-pushed
the
devicetree-mmio
branch
2 times, most recently
from
42b473d
to
a323bbf
Compare
|
|
||
| TEST_CASE(address_translation) | ||
| { | ||
| auto fdt_file = TRY_OR_FAIL(Core::File::open("/usr/Tests/LibDeviceTree/address-translation.dtb"sv, Core::File::OpenMode::Read)); |
There was a problem hiding this comment.
Can you say in the commit message how you created this .dtb file? (Ideally the command line invocation.)
There was a problem hiding this comment.
I manually wrote a dts file and compiled it with dtc -o address-translation.dtb address-translation.dts.
I could add the source file as well, but I wasn't sure if I should do that.
There was a problem hiding this comment.
We also could just depend on dtc for building serenity and compile it from source, but I don't think we want more build dependencies.
There was a problem hiding this comment.
Maybe just put the dtc in to the commit message
spholz
force-pushed
the
devicetree-mmio
branch
from
a323bbf
to
f05a925
Compare
This (confusingly named) function will probably never be used, since you normally should know whether you are looking for a child or property.
This will be necessary in the next commit to avoid a name conflict.
The struct is in the DeviceTree namespace, so the prefix is redundant.
The class is in the DeviceTree namespace, so the prefix is redundant. Calling it a 'View' is also somewhat misleading, as it is not simply a cheap-to-copy reference to the underlying memory but is instead marked as AK_MAKE_NONCOPYABLE.
spholz
force-pushed
the
devicetree-mmio
branch
2 times, most recently
from
92de3d7
to
280a519
Compare
I created this test file by running the following command:
dtc -o address-translation.dtb <<EOF
/dts-v1/;
/ {
compatible = "serenity,address-translation-test";
#address-cells = <2>;
#size-cells = <1>;
soc {
compatible = "simple-bus";
#address-cells = <1>;
#size-cells = <1>;
ranges = <0xa0000000 0xfe 0xd0000000 0x40000000>;
usb@a0010000 {
reg = <0xa0010000 0x100000>;
};
some-bus@b0000000 {
compatible = "simple-bus";
#address-cells = <2>;
#size-cells = <2>;
ranges = <0x02 0x00 0xb0000000 0x00 0x200000>;
leds@200100000 {
reg = <0x02 0x100000 0x00 0x1000>;
};
};
};
};
EOF
spholz
force-pushed
the
devicetree-mmio
branch
from
280a519
to
301babb
Compare
This PR removes 2 hard-coded address offsets and 1 hard-coded interrupt number from the aarch64 code!