From: Claude Code Review Bot <claude-review@example.com>
To: dri-devel-reviews@example.com
Subject: Claude review: drm/imagination: Synchronize interrupts before suspending the GPU
Date: Wed, 11 Mar 2026 13:25:24 +1000 [thread overview]
Message-ID: <review-patch1-20260310-drain-irqs-before-suspend-v1-1-bf4f9ed68e75@imgtec.com> (raw)
In-Reply-To: <20260310-drain-irqs-before-suspend-v1-1-bf4f9ed68e75@imgtec.com>
Patch Review
**Correctness: Good**
The core fix is adding `synchronize_irq()` in `pvr_power_fw_disable()` when called from the runtime PM suspend path (`rpm_suspend=true`). This ensures any in-progress threaded IRQ handler completes before `pvr_fw_stop()` shuts down the firmware and the GPU is powered off.
The removal of `pvr_power_get()`/`pvr_power_put()` from the safety event handling in the threaded IRQ handler is correct and necessary. The commit message explains this well:
1. The threaded IRQ handler is already running because an actual hardware interrupt occurred, so the GPU must be powered on at that point.
2. The `synchronize_irq()` in the suspend path blocks the suspend callback until the threaded handler finishes, so the GPU stays powered throughout the handler's execution.
3. Keeping `pvr_power_get()` in the handler would deadlock: `synchronize_irq()` is called from the runtime PM suspend callback (which holds the PM lock), and `pvr_power_get()` → `pm_runtime_resume_and_get()` requires that same lock.
**Minor observation**: The `rpm_suspend` boolean parameter added to `pvr_power_fw_disable` is a reasonable approach, though an alternative would have been to call `synchronize_irq()` directly in `pvr_power_device_suspend()`. The chosen approach keeps the synchronization close to the firmware stop, which is arguably the right abstraction level.
**Note on the hardirq handler**: The primary hardirq handler (`pvr_device_irq_handler`) also accesses GPU registers via `pvr_fw_irq_pending()` and `pvr_device_safety_irq_pending()`. There is a small theoretical window between `synchronize_irq()` returning and `pvr_fw_stop()` completing where a new hardirq could fire and access registers. Patch 2 addresses this properly.
**Stable backport note**: The Fixes/Cc-stable tags and the below-the-fold explanation about which commit affects which kernel version are well done, making backporting easier for the stable maintainers.
---
Generated by Claude Code Patch Reviewer
next prev parent reply other threads:[~2026-03-11 3:25 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-03-10 11:41 [PATCH 0/2] drm/imagination: Drain interrupts before suspending the GPU Alessio Belle
2026-03-10 11:41 ` [PATCH 1/2] drm/imagination: Synchronize " Alessio Belle
2026-03-11 3:25 ` Claude Code Review Bot [this message]
2026-03-10 11:41 ` [PATCH 2/2] drm/imagination: Disable " Alessio Belle
2026-03-11 3:25 ` Claude review: " Claude Code Review Bot
2026-03-11 3:25 ` Claude review: drm/imagination: Drain " Claude Code Review Bot
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=review-patch1-20260310-drain-irqs-before-suspend-v1-1-bf4f9ed68e75@imgtec.com \
--to=claude-review@example.com \
--cc=dri-devel-reviews@example.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox