Add support for admv1355 - #3576

LLM review

This series adds the ADMV1355 microwave upconverter IIO driver, its DT
binding, a Kconfig.adi imply, and a ZedBoard example devicetree.

Session: 36707494438

eaa815bdf024 - iio: frequency: admv1355: Add support

Switch/case ordering bug: build_llvm_x86_64 fails strict CI with
drivers/iio/frequency/admv1355.c:388:2: error: cannot jump from switch statement to this case label. In admv1355_read_raw(), guard(mutex)()
(a cleanup-attribute variable) is declared directly under
case IIO_CHAN_INFO_PROCESSED: in a switch that also has a default:
label. Jumping to default: bypasses construction of the guard variable,
which C disallows; clang errors on it (and separately warns about it as a
C23 extension). Reproduced locally with clang-19 using the CI flags, and
verified the fix (wrapping the case body in { }) removes both the error
and the C23-extension warning.

Everywhere else in the file, guard(mutex) is taken before entering a
switch, so this is the only instance of the bug.

CI warnings

The checkpatch/coccicheck warnings (macro argument reuse in
ADMV1355_REG_ENUM/ADMV1355_FILTER_ENUM, missing blank lines, mutex
comment, str_true_false() opportunities) are all non-strict/style-only;
the macro arguments are always plain identifiers with no side effects, so
the "argument reuse" warning is a false positive. Not fixed.

The dtc unit_address_vs_reg warnings on clock@0 in the new dts, and
the gcc_fanalyzer warning inside include/linux/cleanup.h, are
pre-existing patterns unrelated to this series (checked against other
zynq-zed*.dts files and the generic guard()/DEFINE_CLASS macro).

Verification data

Datasheet admv1355.md (docling-converted, /tmp/docling.1pa) was used to
verify: the temperature-sensor equation and ADC register sequence in
admv1355_read_temp() against datasheet Equation 3 and the "ADC
Configuration" procedure (register addresses/values match 1:1), and the LO
×4/×3/×1 tunable-filter lookup tables against Table 17 and the
REG0800/0801/0802 bit-field descriptions (all band boundaries and register
codes match exactly). No datasheet-correctness issues found.

make dt_binding_check for adi,admv1355.yaml passes with no errors, and
the example DT builds. The new zynq-zed-adv7511-admv1355.dts is not added
to arch/arm/boot/dts/xilinx/Makefile's dtb-y list, but this matches the
established convention for ADI board dts files in this tree (only upstream
boards are listed there; ADI board dts files are built directly by CI).

Suggested patches

• 0001-fixup-iio-frequency-admv1355-Add-support.patch (eaa815bdf024) —
  wraps the IIO_CHAN_INFO_PROCESSED case body in braces in
  admv1355_read_raw() to fix the clang build failure.

Apply with:

cd path/to/repository
export GITHUB_TOKEN=ghp_***
apply-patches --repo=analogdevicesinc/linux 36707494438

curl -fSsL "https://raw.githubusercontent.com/analogdevicesinc/doctools/refs/heads/main/ci/scripts/apply-patches.sh"      -o ~/.local/bin/apply-patches.sh &&   grep -q "/apply-patches.sh" ~/.bashrc || echo "source ~/.local/bin/apply-patches.sh" >> $_ ; . $_

Hey @danmois, so this is going to be a fun one. For starters this looks to be very upstreamable so I hope that to be the ultimate goal (naturally without the clocking scale out of tree thingy).

Having said the above, all the custom attributes need to have a proper justification for why we need them and have to be documented in Documentation/ABI/testing. For the binary attrs consider using:

https://elixir.bootlin.com/linux/v7.3-rc5/source/drivers/base/firmware_loader/sysfs_upload.c#L294

IIRC @rodrigo455 also used it for loading some metadata to a DDS driver in some operating mode so there's precedence for this kind of stuff.

Last but not least, the driver has 2k+ LOC, consider splitting it into more logical features so it makes it easier to review each patch