Re: [PATCH 3/3] arch/x86: Optionally flush L1D on context switch

[Borislav Petkov]

On Thu, Apr 02, 2020 at 05:24:01PM +1100, Balbir Singh wrote:
> +int enable_l1d_flush_for_task(struct task_struct *tsk)
> +{
> +	struct page *page;
> +	int ret = 0;
> +
> +	if (static_cpu_has(X86_FEATURE_FLUSH_L1D))
> +		goto done;
> +
> +	page = READ_ONCE(l1d_flush_pages);
> +	if (unlikely(!page)) {
> +		mutex_lock(&l1d_flush_mutex);
> +		if (!l1d_flush_pages) {
> +			l1d_flush_pages = alloc_l1d_flush_pages();
> +			if (!l1d_flush_pages)
> +				return -ENOMEM;

Do I see it correctly that you'll return here with l1d_flush_mutex still
held?

> +		}
> +		mutex_unlock(&l1d_flush_mutex);
> +	}
> +	/* I don't think we need to worry about KSM */

Pls use passive voice: no "we" or "I", etc.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette