Skip to content

arm64: dts: exynos: google: zumapro-tegu: Add the NFC controller - #105

Merged
derekross merged 2 commits into
derekross:omamobilefrom
W4HO-ham:tegu-dt-nfc
Nov 1, 2026
Merged

derekross merged 2 commits into
derekross:omamobilefrom
W4HO-ham:tegu-dt-nfc

Conversation

@W4HO-ham

Copy link
Copy Markdown
Contributor

This is the Pixel 9a's NFC in the device tree. It is stacked on the DT series (#80, #81, #85, #87, #90, #96), on #100 (ST54L) and on #104 (the PMU clock pad), so only the last two commits are new:

  • zumapro: #clock-cells = <1> on the PMU (clock 0 is the NFC pad, soc, nfc: the Pixel 9a's NFC clock pad as a PMU clock #104).
  • zumapro-tegu:
    • the ST54L at 0x08 on hsi2c9, st,st54l with the nxp,nxp-nci-i2c fallback;
    • its interrupt gpa7-5, level high, which takes gpa7's eight EINTs (SPI 36 to 43), added the way the board adds gpa4's and gpa6's;
    • RST_N on gph1-6;
    • clocks = <&pmu_system_controller 0>;
    • a hog on gpa0-1, the chip enable that stock's eSE driver raises.

The board DT builds with W=1 (no new warnings) and dt-validate reports nothing for these nodes on today's omamobile. On the 9a, the same nodes in my integration tree probe on nxp-nci, poll, and read a contactless card through neard (#100).

@derekross

Copy link
Copy Markdown
Owner

Thank you. I checked the new nodes against Google's tegu and zumapro DTs: gpa7's SPIs 36-43 are stock's IRQ_ALIVE_EINT26-33, the chip enable on gpa0-1 is stock's ese_chip_enable-gpio (zuma-tegu-ese.dtsi), and RST_N on gph1-6 and the interrupt on gpa7-5 match zuma-tegu-nfc.dtsi. The gate passes on today's omamobile merged with the whole stack, CI's dtbs_check is clean, and the other six Google boards' DTBs come out byte-identical.

One fix: 0666f11232b2 adds two blank lines before the haptics comment (checkpatch: "Please don't use multiple blank lines"). Please drop them.

On merging: this PR carries #100's and #104's commits as well as the board stack. My plan is to merge #100 and #104 on their own first, then the board stack, and then ask you to rebase this onto omamobile so only your two commits are left (#106's ramoops commit on top of them). If you'd rather land it all through #106 at the end, say so and I'll close #100 and #104 as included.

A note for tegu's NFC notes, not for this PR: the hog keeps the whole ST54L powered, eSE included, even with NFC switched off. That's fine for now, but worth writing down with the power numbers.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Dropped the two stray blank lines above the haptics comment (now d8be32c). #106, #119, #122 and #118 are rebased onto it; the DT builds at each tip.

@derekross

Copy link
Copy Markdown
Owner

Thanks, the two blank lines are gone (d8be32c893ce), and the 9a's DTB builds without warnings at this commit.

This PR still carries #100's and #104's commits unchanged, so it lands after those two, which are waiting on their own threads (the exact-build NFC run, and the comment fixes in gs101-pmu.c). Like the rest of the stack, it also needs the rebase onto today's omamobile described on #80. Once #100, #104 and the board stack are in, a rebase leaves only your two device-tree commits here.

@derekross

Copy link
Copy Markdown
Owner

Nothing has changed here since my last look, and the stack still merges cleanly and passes the gate on today's omamobile. What holds it is below it: #96 needs the display and GPU rows from a rebuild with the panel and clk-acpm options on and no blacklist (my comment there, which also notes I've just landed #81 and am adding the panel and thermal drivers to our config fragment). #104 still has the three comment fixes, and I'll come back to you there on how the PMU clock pad should be provided, because my Pixel 7 Pro series drives the same pad registers through clk-exynos-clkout. #131 will land #100's commits. Once those are in, a rebase leaves your two commits here, then #106, #119 and #136 follow in that order.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Rebased onto #96's new head. The ST54L pair is in omamobile now and dropped out, and the branch now carries #104's current four commits (with your comment fixes) under the two of its own: the PMU clocks line and the NFC controller. Compiles; checkpatch clean.

@derekross

Copy link
Copy Markdown
Owner

Thanks for the rebase. One correction: the four #104 commits here are the same patches as #104's head from 10-28 (gs101-pmu.c is unchanged), so the three comment fixes aren't in yet. Since #104 is moving to clk-exynos-clkout, this PR will rebase onto that rework, and only your two DT commits will remain.

@W4HO-ham

W4HO-ham commented Nov 1, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto #104's clock-out rework, on #96's current head, so only my two DT commits remain on top of #104 ("The PMU provides clocks" and the NFC controller). The PMU node keeps #clock-cells = <1> and the NFC node clocks = <&pmu_system_controller 0>, which is what the clock-out pad variant serves, with no clock-names. This is the exact head booted for #104's test, with the result posted there: nfc0 polling with the clock enabled and 0 I2C errors.

@derekross

Copy link
Copy Markdown
Owner

Thanks, @W4HO-ham. This is now just your two device-tree commits on top of #104's rework and #96, and the tree is exactly the one you booted. Your release's hash, b45a0d400e49, has the same tree as this head.

What I checked:

  • The PMU node keeps #clock-cells = <1>, and the NFC node takes clocks = <&pmu_system_controller 0> with no clock-names. That's the same layout my gs201 commit uses (#clock-cells on the PMU node itself), so my "a clock-controller node" on soc, nfc: the Pixel 9a's NFC clock pad as a PMU clock #104 was loose wording on my part. What you did is right.
  • The 9a's DTB at this head passes dt-validate with no messages against this branch's bindings, and the other seven Google DTBs are byte-identical to the base. The gate passes merged onto today's omamobile (W=1, no new device-tree warnings), and CI is green.
  • Your boot shows pmu-clkout0 enabled with 1-0008 as its consumer, nfc0 polling with all six protocols, and no nxp-nci I2C errors. The I2C side working is itself evidence that EXTPAD_EN is set, since without it the chip doesn't ACK.

Two small things in the messages before it lands:

  1. "The PMU provides clocks" says "The zuma PMU". The file is zumapro.dtsi, and the binding now says Tensor G4, so please say Tensor G4 (or zumapro) here too.
  2. "Add the NFC controller" says the controller "polls and reads a contactless card (ISO-DEP over NFC-A)". That card read was the 10-28 run, with the old gs101-pmu provider. Today's boot of this tree shows polling, but no card. The register bit is the same, so I expect it to hold. Still, the message should describe this tree. When you're next at the palmtop, one card tap on this head would cover it (nfctool -l and the neard lines). If that's not convenient, say in the message that the read was on the earlier provider.

On order: this lands after #96 and #104. Once both are in, a git rebase omamobile leaves your two commits. It and #119 add neighbouring blocks (&gpa7 here, &gpa8 there), so whichever goes second picks up a small context conflict. The resolution keeps both, &gpa7 first, and I've put the exact hunk on #119.

@W4HO-ham

W4HO-ham commented Nov 1, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto today's omamobile (with #96); details on #119.

@derekross

Copy link
Copy Markdown
Owner

Thanks for the rebase, @W4HO-ham. All nine commits are the same patches as before, so the tree is still the one you booted, and the gate passes merged onto today's omamobile with CI green.

The rebase went up a few minutes after my last comment, so I think the two message points crossed it. Both are still open in the head:

  1. "zumapro: The PMU provides clocks" still begins "The zuma PMU provides the NFC controller's clock pad". Please say Tensor G4 (or zumapro), as the binding now does.
  2. "zumapro-tegu: Add the NFC controller" still says it "polls and reads a contactless card (ISO-DEP over NFC-A)". That card read was on the old provider. Either one card tap on this tree (nfctool -l and the neard lines), or a sentence saying the read was on the earlier provider.

The order is unchanged: this lands after #104, and whichever of this and #161 goes second takes the &gpa7/&gpa8 hunk I put on #119. #161 is merged now, so that's this one. When #104 is in, a git rebase omamobile leaves your two commits.

@derekross

Copy link
Copy Markdown
Owner

Thank you, @W4HO-ham. Both message points are done: "The PMU provides clocks" now says Tensor G4 (zumapro), and "Add the NFC controller" says the card read was on the earlier version of the series, with the other provider. The tree is the same as before.

#104 has landed, so this needs one more git rebase omamobile. Git drops the seven #104 commits on its own (they're the same patches as the ones on omamobile). "Add the NFC controller" then hits the context conflict next to &gpa6, now that #161's &gpa8 block is in. The resolution is the hunk I put on #119: keep both blocks, &gpa7 first, then &gpa8. Nothing else in that commit changes.

I tried that rebase locally: the two commits, with the conflict resolved that way, pass the gate merged onto today's omamobile (both signed off, checkpatch --strict with nothing to read, eight device trees with no new W=1 warnings). They also merge cleanly with #136. CI hasn't run on your last push because the branch no longer merges onto omamobile. It will run once you push the rebase, and then I'll merge.

The Tensor G4 (zumapro) PMU provides the NFC controller's clock pad as
clock 0; give its node #clock-cells.

Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
The Pixel 9a's ST54L sits at 0x08 on hsi2c9, its interrupt on gpa7-5
(level high), its RST_N on gph1-6, and its reference clock from the PMU's
clock pad. Its chip enable (gpa0-1) powers the whole chip and is held
high by a hog: stock's eSE driver raises it, and nothing here drives the
eSE.

Tested on the Pixel 9a: on this tree the controller probes on nxp-nci
and neard polls with the clock enabled, with no I2C errors. The read of
a contactless card (ISO-DEP over NFC-A) was on the earlier version of
the series, which took the clock from a different provider.

Link: https://android.googlesource.com/kernel/devices/google/tegu/+/6276884d9940694b25a6934671918ec8ac412ee8/dts/zuma-tegu-nfc.dtsi
Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
@W4HO-ham

W4HO-ham commented Nov 1, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto omamobile (8e84093) now that #104's commits are in, which leaves my two device-tree commits (head 12d4f25). As you said, this one takes the &gpa7/&gpa8 hunk: &gpa7 (the NFC interrupt, SPI 36-43) now sits right before #161's &gpa8, and both are as before. The 9a's DTB builds with no dtc warnings, and checkpatch is clean on both. The messages are the reworded ones from earlier today.

@derekross

Copy link
Copy Markdown
Owner

Thank you, @W4HO-ham, for the quick rebase. The two commits are the same patches as before. The only change is the context next to &gpa8: &gpa7 now sits right before #161's block, and both blocks are as they were. So the tree is still the one you booted.

On today's omamobile (b39c04d) the gate passes: a clean merge, both commits signed off, checkpatch --strict with nothing to read, and eight device trees with no new W=1 warnings. CI is green on this head too. I also ran dt-validate on the merged Pixel 9a DTB against the merged bindings, and it's clean, as on the base. It merges cleanly with the other open 9a PRs (#55, #99, #125, #178).

I'll rebase-merge it. Thank you for seeing this one through #104's rework.

@derekross
derekross merged commit b7de5f1 into derekross:omamobile Nov 1, 2026
1 check passed
@derekross

Copy link
Copy Markdown
Owner

Merged. Thanks, @W4HO-ham.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants