From mboxrd@z Thu Jan  1 00:00:00 1970
From: Claude Code Review Bot <claude-review@example.com>
To: dri-devel-reviews@example.com
Subject: Claude review: drm/display/dp: Export function to wake the sink AUX_CH
Date: Tue, 10 Mar 2026 12:41:16 +1000
Message-ID: <review-patch1-20260309-dp_aux_timeout-v1-1-08c610a63a84@intel.com>
In-Reply-To: <20260309-dp_aux_timeout-v1-1-08c610a63a84@intel.com>
References: <20260309-dp_aux_timeout-v1-0-08c610a63a84@intel.com> <20260309-dp_aux_timeout-v1-1-08c610a63a84@intel.com>
X-Mailer: Claude Code Patch Reviewer
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0

Patch Review

**`poll_timeout_us` misuse:**

```c
ret =3D poll_timeout_us(ret =3D drm_dp_dpcd_readb(aux, DP_SET_POWER, &value=
),
                      ret > 0,
                      1000, wake_retry * 1000, true);
```

The variable `ret` is used both as the operation result (inside the macro e=
xpansion) AND as the macro's return value. After the macro completes, `ret`=
 holds the macro's return value (0 or `-ETIMEDOUT`), not the `drm_dp_dpcd_r=
eadb` return value. This means the `ret < 0` check below is only checking f=
or `-ETIMEDOUT` from the poll, not a DPCD read failure. This is confusing a=
nd fragile =E2=80=94 use a separate variable for the poll result.

**Condition logic concern:**

```c
if (value =3D=3D DP_SET_POWER_D3 || ret < 0) {
```

If the poll times out (ret =3D=3D -ETIMEDOUT), `value` might contain stale =
or partial data from a failed read. The function proceeds to write to `DP_S=
ET_POWER` without checking whether DPCD writes actually succeed =E2=80=94 n=
o error checking on either `drm_dp_dpcd_writeb` call.

**Wake sequence concern:**

```c
drm_dp_dpcd_writeb(aux, DP_SET_POWER, DP_SET_POWER_D0);
fsleep(1000);
drm_dp_dpcd_writeb(aux, DP_SET_POWER, DP_SET_POWER_D3_AUX_ON);
```

Setting to D0 and then immediately to D3_AUX_ON seems odd. The cover letter=
 and spec reference (Section 2.3.4) say the goal is to wake the sink for AU=
X communication, but D3_AUX_ON is still a low-power state. If the caller is=
 about to retry an AUX transaction, why not leave it in D0? The intent shou=
ld be clarified.

**Missing `#include`:** The function uses `poll_timeout_us` from `<linux/io=
poll.h>` and `fsleep`. Ensure the drm_dp_helper.c file includes the necessa=
ry headers.

**Naming:** `drm_dp_wake_sink` is generic but it specifically deals with AU=
X channel power states. A name like `drm_dp_aux_wake_sink` might be more pr=
ecise.

**Documentation:** The kerneldoc is incomplete =E2=80=94 it should describe=
 the behavior (what it does on success/failure, return value semantics =E2=
=80=94 though it's void).

---

---
Generated by Claude Code Patch Reviewer