Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <fd5dd250-b01e-fd00-1419-88398cb13347@loongson.cn>
Date:   Thu, 22 Jun 2023 01:21:44 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.11.0
Subject: Re: [PATCH v10 07/11] drm/etnaviv: Add support for the dma coherent
 device
Content-Language: en-US
To:     Lucas Stach <l.stach@pengutronix.de>,
        Sui Jingfeng <18949883232@163.com>,
        Russell King <linux+etnaviv@armlinux.org.uk>,
        Christian Gmeiner <christian.gmeiner@gmail.com>,
        David Airlie <airlied@gmail.com>,
        Daniel Vetter <daniel@ffwll.ch>
Cc:     linux-kernel@vger.kernel.org, dri-devel@lists.freedesktop.org,
        etnaviv@lists.freedesktop.org,
        Philipp Zabel <p.zabel@pengutronix.de>,
        Bjorn Helgaas <bhelgaas@google.com>
References: <20230620094716.2231414-1-18949883232@163.com>
 <20230620094716.2231414-8-18949883232@163.com>
 <8f74f0962c8bab6c832919a5340667c54e1a7ddc.camel@pengutronix.de>
 <66fc74ae-299c-a5de-9cfb-07ae24fb3f07@loongson.cn>
 <8212078bd56c54ce508205eae0ed0b69e78d4c38.camel@pengutronix.de>
From:   Sui Jingfeng <suijingfeng@loongson.cn>
Organization: Loongson
In-Reply-To: <8212078bd56c54ce508205eae0ed0b69e78d4c38.camel@pengutronix.de>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

Hi,

On 2023/6/21 23:58, Lucas Stach wrote:
> Am Mittwoch, dem 21.06.2023 um 23:30 +0800 schrieb Sui Jingfeng:
>> Hi,
>>
>> On 2023/6/21 18:00, Lucas Stach wrote:
>>>>    		dma_sync_sgtable_for_cpu(dev->dev, etnaviv_obj->sgt,
>>>>    					 etnaviv_op_to_dma_dir(op));
>>>>    		etnaviv_obj->last_cpu_prep_op = op;
>>>> @@ -408,8 +421,9 @@ int etnaviv_gem_cpu_fini(struct drm_gem_object *obj)
>>>>    {
>>>>    	struct drm_device *dev = obj->dev;
>>>>    	struct etnaviv_gem_object *etnaviv_obj = to_etnaviv_bo(obj);
>>>> +	struct etnaviv_drm_private *priv = dev->dev_private;
>>>>    
>>>> -	if (etnaviv_obj->flags & ETNA_BO_CACHED) {
>>>> +	if (!priv->dma_coherent && etnaviv_obj->flags & ETNA_BO_CACHED) {
>>>>    		/* fini without a prep is almost certainly a userspace error */
>>>>    		WARN_ON(etnaviv_obj->last_cpu_prep_op == 0);
>>>>    		dma_sync_sgtable_for_device(dev->dev, etnaviv_obj->sgt,
>>>> diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c b/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c
>>>> index 3524b5811682..754126992264 100644
>>>> --- a/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c
>>>> +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c
>>>> @@ -112,11 +112,16 @@ static const struct etnaviv_gem_ops etnaviv_gem_prime_ops = {
>>>>    struct drm_gem_object *etnaviv_gem_prime_import_sg_table(struct drm_device *dev,
>>>>    	struct dma_buf_attachment *attach, struct sg_table *sgt)
>>>>    {
>>>> +	struct etnaviv_drm_private *priv = dev->dev_private;
>>>>    	struct etnaviv_gem_object *etnaviv_obj;
>>>>    	size_t size = PAGE_ALIGN(attach->dmabuf->size);
>>>> +	u32 cache_flags = ETNA_BO_WC;
>>>>    	int ret, npages;
>>>>    
>>>> -	ret = etnaviv_gem_new_private(dev, size, ETNA_BO_WC,
>>>> +	if (priv->dma_coherent)
>>>> +		cache_flags = ETNA_BO_CACHED;
>>>> +
>>> Drop this change. Instead etnaviv_gem_new_impl() should do the upgrade
>>> from WC to CACHED as necessary by adding something like this:
>> I understand you are a profession person in vivante GPU driver domain.
>>
>> I respect you reviews and instruction.
>>
>> But, I'm really reluctant to agree with this, is there any space to
>> negotiate?
>>
>>> /*
>>>    * Upgrade WC to CACHED when the device is hardware coherent and the
>>>    * platform doesn't allow mixing cached and writecombined mappings to
>>>    * the same memory area.
>>>    */
>>> if ((flags & ETNA_BO_CACHE_MASK) == ETNA_BO_WC &&
>>>       dev_is_dma_coherent(dev) && !drm_arch_can_wc_memory())
>>>           flags = (flags & ~ETNA_BO_CACHE_MASK) & ETNA_BO_CACHED;
>> This is policy, not a mechanism.
>>
>> Using what cache property is a user-space program's choice.
>>
>> While you are override the WC with CACHED mapping. This is not correct
>> in the concept!
>>
> Please explain why you think that this isn't correct.

Again,

this is user-space things!

this is user-space things!

this is user-space things!

I have explained several times.

made the decision for the user-space program is wrong.


> It allows
> userspace to use WC mappings that would potentially cause loss of
> coherency between CPU and GPU, which isn't acceptable.

Before made the WC works correctly,  you need the developing environment.

userspace program can tune the BO cache mapping easily.

Either environment or supply a conf file.


While with your implement, we don't have the opportunity to do debugging 
and the development.

The kernel space is writing hard-code.


> Regards,
> Lucas

-- 
Jingfeng