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: accel/amdxdna: Preserve user address when PASID is disabled
Date: Thu, 04 Jun 2026 13:24:31 +1000
Message-ID: <review-overall-20260602040624.2206774-1-lizhi.hou@amd.com>
In-Reply-To: <20260602040624.2206774-1-lizhi.hou@amd.com>
References: <20260602040624.2206774-1-lizhi.hou@amd.com>
X-Mailer: Claude Code Patch Reviewer
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
MIME-Version: 1.0

Overall Series Review

Subject: accel/amdxdna: Preserve user address when PASID is disabled
Author: Lizhi Hou <lizhi.hou@amd.com>
Patches: 3
Reviewed: 2026-06-04T13:24:31.255992

---

This is a single-patch fix for the `accel/amdxdna` driver. The bug is clear: when PASID is disabled, `amdxdna_hmm_register()` returns early without setting `abo->mem.uva`, so it stays at `AMDXDNA_INVALID_ADDR` (its initialization value from `amdxdna_gem_create_obj()` at line 178). Later, `amdxdna_init_dev_bo()` calls `amdxdna_gem_uva(heap)` on each heap BO and checks:

```c
heap_addr = amdxdna_gem_uva(heap);
if (heap_addr == AMDXDNA_INVALID_ADDR) {
    XDNA_ERR(xdna, "Heap %d is not mapped", heap_id);
    return -EAGAIN;
}
```

This validation always fails in the no-PASID path because `uva` was never populated, even though the BO was successfully mapped. The fix is correct and minimal.

One minor observation: there is a behavioral asymmetry between the PASID-on and PASID-off paths after this patch that is worth noting but not blocking.

---

---
Generated by Claude Code Patch Reviewer