mirror of
https://github.com/torvalds/linux.git
synced 2026-09-29 21:04:02 +02:00
8e5a6599a3
Runyu Xiao <runyu.xiao@seu.edu.cn> says: Both cs35l33 and cs35l34 can enter runtime suspend while their threaded IRQ handlers are still reachable. The suspend path then switches the driver into regcache cache-only mode and powers the codec down, while the IRQ thread still expects live status-register access. This issue was found by our static analysis tool and manually audited on Linux v6.18.21. Directed QEMU no-device validation further showed that, after runtime_suspend() completed, the real threaded handlers could still be injected and would continue past volatile regmap read failures into their release/update paths. This series keeps the fix on the suspend actor and applies the same runtime-suspend/threaded-IRQ ordering repair boundary to both drivers: - track whether request_threaded_irq() actually succeeded - disable_irq() before cache_only/power-off in runtime suspend - enable_irq() only after regcache_sync() in runtime resume That drains any in-flight threaded handler and blocks new IRQ handling while the codec is suspended, without mixing in larger IRQ-thread defensive cleanups. Build-tested by compiling cs35l33.o and cs35l34.o. No cs35l33/cs35l34 hardware was available for end-to-end runtime testing. Link: https://patch.msgid.link/20260611161553.3378721-1-runyu.xiao@seu.edu.cn |
||
|---|---|---|
| .. | ||
| acpi | ||
| asm-generic | ||
| clocksource | ||
| crypto | ||
| cxl | ||
| drm | ||
| dt-bindings | ||
| hyperv | ||
| keys | ||
| kunit | ||
| kvm | ||
| linux | ||
| math-emu | ||
| media | ||
| memory | ||
| misc | ||
| net | ||
| pcmcia | ||
| ras | ||
| rdma | ||
| rv | ||
| scsi | ||
| soc | ||
| sound | ||
| target | ||
| trace | ||
| uapi | ||
| ufs | ||
| vdso | ||
| video | ||
| xen | ||
| Kbuild | ||