Re: [PATCH] iommu/exynos: Initialize lock before requesting IRQ

[Runyu Xiao]

Hello Robin and Marek,

Thanks for the reviews.

I agree that the current patch is incomplete. The interrupt handler does
not only access data->lock; it also depends on the clocks, variant data,
device pointers, and the SYSMMU hardware state, which are not fully
initialized at the current devm_request_irq() point.

Moving only spin_lock_init() before devm_request_irq() therefore does not
make the handler safe. I will withdraw this patch and will not send a v2
of the current change.

If this issue is pursued further, I will review the complete probe ordering
and the SYSMMU hardware disable/reset sequence before preparing a new patch.

Please disregard the submitted patch.

Regards,
Runyu