diff options
| author | Chancel Liu <chancel.liu@nxp.com> | 2026-09-23 15:13:02 +0900 |
|---|---|---|
| committer | Mark Brown <broonie@kernel.org> | 2026-09-23 14:23:01 +0200 |
| commit | 57d1e55c7b5a66a8642e3c944d46e3d3bafb28c8 (patch) | |
| tree | 9eb7470b35104aee1145846443f87ecb9317d83f /include/linux/timerqueue.h | |
| parent | 5ef6e7ff52db85556ef1fcca393366dfdd6452f9 (diff) | |
ASoC: codecs: rt5677-spi: Free the DSP context on component remove
component->dev is the underlying SPI bus device, whose devres lifetime
follows the physical devices probe/remove rather than the ASoC cards
bind/unbind. The rt5677_dsp context allocated in the component probe with
devm_kzalloc(component->dev, ...) therefore leaks on every card
bind/unbind cycle, and the delayed work initialised there is never
cancelled on unbind.
Allocate the context with kzalloc() and add a component .remove callback
that cancels the copy work, destroys the mutex and frees the context.
The exported rt5677_spi_hotword_detected(), called from the rt5677 codec
interrupt path, fetches the context via dev_get_drvdata(&g_spi->dev) and
dereferences it. Freeing the context on remove would let that callback
run concurrently and touch freed memory. Add a dsp_lock that serializes
the fetch/use in rt5677_spi_hotword_detected() against clearing the
drvdata in the remove callback, so once remove has cleared the pointer
the callback either observes NULL and bails out or has already finished.
Signed-off-by: Chancel Liu <chancel.liu@nxp.com>
Link: https://patch.msgid.link/20260923061305.4041197-19-chancel.liu@oss.nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Diffstat (limited to 'include/linux/timerqueue.h')
0 files changed, 0 insertions, 0 deletions
