[PATCH v3 12/33] iommu/mediatek: Always tlb_flush_all when each PM resume

Dafna Hirschfeld dafna.hirschfeld at collabora.com
Tue Nov 9 21:29:27 PST 2021



On 10.11.21 04:20, Yong Wu wrote:
> On Tue, 2021-11-09 at 14:21 +0200, Dafna Hirschfeld wrote:
>> Hi
>> This patch is needed in order to update the tlb when a device is
>> powered on.
>> Could you send this patch alone without the whole series so it get
>> accepted easier?
> 
> Which SoC are you testing on? In previous SoC, the IOMMU HW don't have
> power-domain, and we have a "has_pm"[1] in the tlb function for that
> case. The "has_pm" should be always 0 for the previous SoC like mt8173,
> it should always tlb synchronize.
> 
> thus, Could you help share more about your issue? In which case it lack
> the necessary tlb operation. At least, We need confirm if it needs a
> "Fixes" tags if sending this patch alone.

Hi,
I work with the mtk-vcodec driver on mt8173. As you wrote, the iommu doesn't
have a power-domain and so when allocating buffers before the device is powered
on, there is the warning
"Partial TLB flush timed out, falling back to full flush"
flooding the log buf.

Sebastian Reichel suggested to remove the 'if(has_pm)' check to avoid this warning,
and avoid flushing the tlb if the device is off:

[1] http://ix.io/3Eyr

This fixes the warning, but then the tlb is not flushed in sync,
Therefore the tlb should be flushed when the device is resumed.

So the two patches (the one suggested in the link [1] and this patch)
should be sent together as a 2-patch series.

Thanks,
Dafna

> 
> Thanks.
> 
> [1]
> https://elixir.bootlin.com/linux/v5.15/source/drivers/iommu/mtk_iommu.c#L236
> 
>> I can resend the patch on your behalf if you want.
>>
>> Thanks,
>> Dafna
>>
>> On 23.09.21 14:58, Yong Wu wrote:
>>> Prepare for 2 HWs that sharing pgtable in different power-domains.
>>>
>>> When there are 2 M4U HWs, it may has problem in the flush_range in
>>> which
>>> we get the pm_status via the m4u dev, BUT that function don't
>>> reflect the
>>> real power-domain status of the HW since there may be other HW also
>>> use
>>> that power-domain.
>>>
>>> The function dma_alloc_attrs help allocate the iommu buffer which
>>> need the corresponding power domain since tlb flush is needed when
>>> preparing iova. BUT this function only is for allocating buffer,
>>> we have no good reason to request the user always call
>>> pm_runtime_get
>>> before calling dma_alloc_xxx. Therefore, we add a tlb_flush_all
>>> in the pm_runtime_resume to make sure the tlb always is clean.
>>>
>>> Another solution is always call pm_runtime_get in the
>>> tlb_flush_range.
>>> This will trigger pm runtime resume/backup so often when the iommu
>>> power is not active at some time(means user don't call
>>> pm_runtime_get
>>> before calling dma_alloc_xxx), This may cause the performance drop.
>>> thus we don't use this.
>>>
>>> In other case, the iommu's power should always be active via device
>>> link with smi.
>>>
>>> The previous SoC don't have PM except mt8192. the mt8192 IOMMU is
>>> display's
>>> power-domain which nearly always is enabled. thus no need fix tags
>>> here.
>>> Prepare for mt8195.
>>>
>>> Signed-off-by: Yong Wu <yong.wu at mediatek.com>
>>> ---
>>>    drivers/iommu/mtk_iommu.c | 11 +++++++++++
>>>    1 file changed, 11 insertions(+)
>>>
>>> diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
>>> index 44cf5547d084..e9e94944ed91 100644
>>> --- a/drivers/iommu/mtk_iommu.c
>>> +++ b/drivers/iommu/mtk_iommu.c
>>> @@ -984,6 +984,17 @@ static int __maybe_unused
>>> mtk_iommu_runtime_resume(struct device *dev)
>>>    		return ret;
>>>    	}
>>>    
>>> +	/*
>>> +	 * Users may allocate dma buffer before they call
>>> pm_runtime_get, then
>>> +	 * it will lack the necessary tlb flush.
>>> +	 *
>>> +	 * We have no good reason to request the users always call
>>> dma_alloc_xx
>>> +	 * after pm_runtime_get_sync.
>>> +	 *
>>> +	 * Thus, Make sure the tlb always is clean after each PM
>>> resume.
>>> +	 */
>>> +	mtk_iommu_tlb_do_flush_all(data);
>>> +
>>>    	/*
>>>    	 * Uppon first resume, only enable the clk and return, since
>>> the values of the
>>>    	 * registers are not yet set.
>>>



More information about the linux-arm-kernel mailing list