linux-kernel - Re: [PATCH v3 13/33] iommu/mediatek: Remove the power status checking in tlb flush all

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <6cff0b97-b861-e02d-e76f-2510c962c452@collabora.com>
Date:   Fri, 22 Oct 2021 16:03:19 +0200
From:   Dafna Hirschfeld <dafna.hirschfeld@...labora.com>
To:     Yong Wu <yong.wu@...iatek.com>, Joerg Roedel <joro@...tes.org>,
        Rob Herring <robh+dt@...nel.org>,
        Matthias Brugger <matthias.bgg@...il.com>,
        Will Deacon <will@...nel.org>,
        Robin Murphy <robin.murphy@....com>
Cc:     Krzysztof Kozlowski <krzysztof.kozlowski@...onical.com>,
        Tomasz Figa <tfiga@...omium.org>,
        linux-mediatek@...ts.infradead.org, srv_heupstream@...iatek.com,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org,
        iommu@...ts.linux-foundation.org,
        Hsin-Yi Wang <hsinyi@...omium.org>, youlin.pei@...iatek.com,
        anan.sun@...iatek.com, chao.hao@...iatek.com,
        yen-chang.chen@...iatek.com,
        Collabora Kernel ML <kernel@...labora.com>
Subject: Re: [PATCH v3 13/33] iommu/mediatek: Remove the power status checking
 in tlb flush all

Hi


On 23.09.21 13:58, Yong Wu wrote:
> To simplify the code, Remove the power status checking in the
> tlb_flush_all, remove this:
>     if (pm_runtime_get_if_in_use(data->dev) <= 0)
> 	    continue;
> 
> After this patch, the mtk_iommu_tlb_flush_all will be called from
> a) isr
> b) pm runtime resume callback
> c) tlb flush range fail case
> d) iommu_create_device_direct_mappings
>     -> iommu_flush_iotlb_all
> In first three cases, the power and clock always are enabled; d) is direct

Regarding case "c) tlb flush range fail case", I found out that this often happens
when the iommu is used while it is runtime suspended. For example the mtk-vcodec
encoder driver calls "pm_runtime_resume_and_get" only when it starts streaming but
buffers allocation is done in 'v4l2_reqbufs' before "pm_runtime_resume_and_get" is called
and then I see the warning "Partial TLB flush timed out, falling back to full flush"
I am not sure how to fix that issue, but it seems that case 'c)' might indicate that
power and clock are actually not enabled.

> mapping, the tlb flush is unnecessay since we already have tlb_flush_all
> in the pm_runtime_resume callback. When the iommu's power status is
> changed to active, the tlb always is clean.
> 
> In addition, there still are 2 reasons that don't add PM status checking
> in the tlb flush all:
> a) Write tlb flush all register also is ok even though the HW has no
> power and clocks. Write ignore.
> b) pm_runtime_get_if_in_use(m4udev) is 0 when the tlb_flush_all
> is called frm pm_runtime_resume cb. From this point, we can not add
> this code above in this tlb_flush_all.
> 
> Signed-off-by: Yong Wu <yong.wu@...iatek.com>
> ---
>   drivers/iommu/mtk_iommu.c | 20 +++++++-------------
>   1 file changed, 7 insertions(+), 13 deletions(-)
> 
> diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
> index e9e94944ed91..4a33b6c6b1db 100644
> --- a/drivers/iommu/mtk_iommu.c
> +++ b/drivers/iommu/mtk_iommu.c
> @@ -204,10 +204,14 @@ static struct mtk_iommu_domain *to_mtk_domain(struct iommu_domain *dom)
>   	return container_of(dom, struct mtk_iommu_domain, domain);
>   }
>   
> -static void mtk_iommu_tlb_do_flush_all(struct mtk_iommu_data *data)
> +static void mtk_iommu_tlb_flush_all(struct mtk_iommu_data *data)
>   {
>   	unsigned long flags;
>   
> +	/*
> +	 * No need get power status since the HW PM status nearly is active
> +	 * when entering here.
> +	 */
>   	spin_lock_irqsave(&data->tlb_lock, flags);
>   	writel_relaxed(F_INVLD_EN1 | F_INVLD_EN0,
>   		       data->base + data->plat_data->inv_sel_reg);
> @@ -216,16 +220,6 @@ static void mtk_iommu_tlb_do_flush_all(struct mtk_iommu_data *data)
>   	spin_unlock_irqrestore(&data->tlb_lock, flags);
>   }
>   
> -static void mtk_iommu_tlb_flush_all(struct mtk_iommu_data *data)
> -{
> -	if (pm_runtime_get_if_in_use(data->dev) <= 0)
> -		return;
> -
> -	mtk_iommu_tlb_do_flush_all(data);
> -
> -	pm_runtime_put(data->dev);
> -}
> -
>   static void mtk_iommu_tlb_flush_range_sync(unsigned long iova, size_t size,
>   					   struct mtk_iommu_data *data)
>   {
> @@ -263,7 +257,7 @@ static void mtk_iommu_tlb_flush_range_sync(unsigned long iova, size_t size,
>   		if (ret) {
>   			dev_warn(data->dev,
>   				 "Partial TLB flush timed out, falling back to full flush\n");
> -			mtk_iommu_tlb_do_flush_all(data);
> +			mtk_iommu_tlb_flush_all(data);
>   		}
>   
>   		if (has_pm)
> @@ -993,7 +987,7 @@ static int __maybe_unused mtk_iommu_runtime_resume(struct device *dev)
>   	 *
>   	 * Thus, Make sure the tlb always is clean after each PM resume.
>   	 */
> -	mtk_iommu_tlb_do_flush_all(data);
> +	mtk_iommu_tlb_flush_all(data);
>   
>   	/*
>   	 * Uppon first resume, only enable the clk and return, since the values of the
>