[Intel-gfx] [PATCH] drm/i915/gtt: Avoid overflowing the WC stash

[Matthew Auld]

On 29/05/2019 10:34, Chris Wilson wrote:
> An interesting issue cropped with making the pagetables be allocated and
> freed concurrently (i.e. removing their grandeous struct_mutex guard)
> was that we would overflow the page stash. This happens when we have
> multiple allocators grabbing WC pages such that we fill the vm's local
> page stash and then when we free another page, the page stash is already
> full and we overflow.
> 
> The fix is quite simple to check for a full page stash before adding
> another. This results in us keeping a vm local page stash around for
> much longer which is both a blessing and a curse.
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Matthew Auld <matthew.auld at intel.com>
> Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld at intel.com>

> ---
>   drivers/gpu/drm/i915/i915_gem_gtt.c | 10 ++++++----
>   1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index 7496cce0d798..2d7a968d4fd5 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -341,11 +341,11 @@ static struct page *stash_pop_page(struct pagestash *stash)
>   
>   static void stash_push_pagevec(struct pagestash *stash, struct pagevec *pvec)
>   {
> -	int nr;
> +	unsigned int nr;
>   
>   	spin_lock_nested(&stash->lock, SINGLE_DEPTH_NESTING);
>   
> -	nr = min_t(int, pvec->nr, pagevec_space(&stash->pvec));
> +	nr = min_t(typeof(nr), pvec->nr, pagevec_space(&stash->pvec));
>   	memcpy(stash->pvec.pages + stash->pvec.nr,
>   	       pvec->pages + pvec->nr - nr,
>   	       sizeof(pvec->pages[0]) * nr);
> @@ -399,7 +399,8 @@ static struct page *vm_alloc_page(struct i915_address_space *vm, gfp_t gfp)
>   		page = stack.pages[--stack.nr];
>   
>   		/* Merge spare WC pages to the global stash */
> -		stash_push_pagevec(&vm->i915->mm.wc_stash, &stack);
> +		if (stack.nr)
> +			stash_push_pagevec(&vm->i915->mm.wc_stash, &stack);
>   
>   		/* Push any surplus WC pages onto the local VM stash */
>   		if (stack.nr)
> @@ -469,8 +470,9 @@ static void vm_free_page(struct i915_address_space *vm, struct page *page)
>   	 */
>   	might_sleep();
>   	spin_lock(&vm->free_pages.lock);
> -	if (!pagevec_add(&vm->free_pages.pvec, page))
> +	if (!pagevec_space(&vm->free_pages.pvec))
>   		vm_free_pages_release(vm, false);
> +	pagevec_add(&vm->free_pages.pvec, page);
>   	spin_unlock(&vm->free_pages.lock);
>   }
>   
>