Re: [PATCH v2 1/8] bus: mhi: core: Cache intmod from mhi event to mhi channel

[Jeffrey Hugo]

On 4/27/2020 8:59 PM, Bhaumik Bhatt wrote:
> From: Hemant Kumar <hemantk@xxxxxxxxxxxxxx>
>
> Driver is using zero initialized intmod value from mhi channel when
> configuring TRE for bei field. This prevents interrupt moderation to
> take effect in case it is supported by an event ring. Fix this by
> copying intmod value from associated event ring to mhi channel upon
> registering mhi controller.
>
> Signed-off-by: Hemant Kumar <hemantk@xxxxxxxxxxxxxx>
> Signed-off-by: Bhaumik Bhatt <bbhatt@xxxxxxxxxxxxxx>
> ---
>   drivers/bus/mhi/core/init.c | 4 ++++
>   drivers/bus/mhi/core/main.c | 9 ++++++---
>   2 files changed, 10 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
> index b38359c..4dc7f22 100644
> --- a/drivers/bus/mhi/core/init.c
> +++ b/drivers/bus/mhi/core/init.c
> @@ -864,6 +864,10 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
>   		mutex_init(&mhi_chan->mutex);
>   		init_completion(&mhi_chan->completion);
>   		rwlock_init(&mhi_chan->lock);
> +
> +		/* used in setting bei field of TRE */
> +		mhi_event = &mhi_cntrl->mhi_event[mhi_chan->er_index];
> +		mhi_chan->intmod = mhi_event->intmod;
>   	}
>   
>   	if (mhi_cntrl->bounce_buf) {
> diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
> index eb4256b..23154f1 100644
> --- a/drivers/bus/mhi/core/main.c
> +++ b/drivers/bus/mhi/core/main.c
> @@ -929,7 +929,7 @@ int mhi_queue_skb(struct mhi_device *mhi_dev, enum dma_data_direction dir,
>   	struct mhi_ring *buf_ring = &mhi_chan->buf_ring;
>   	struct mhi_buf_info *buf_info;
>   	struct mhi_tre *mhi_tre;
> -	int ret;
> +	int ret, bei;
>   
>   	/* If MHI host pre-allocates buffers then client drivers cannot queue */
>   	if (mhi_chan->pre_alloc)
> @@ -966,10 +966,11 @@ int mhi_queue_skb(struct mhi_device *mhi_dev, enum dma_data_direction dir,
>   		goto map_error;
>   
>   	mhi_tre = tre_ring->wp;
> +	bei = !!(mhi_chan->intmod);
>   
>   	mhi_tre->ptr = MHI_TRE_DATA_PTR(buf_info->p_addr);
>   	mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(buf_info->len);
> -	mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(1, 1, 0, 0);
> +	mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, 1, 0, 0);

Why are we not using mhi_gen_tre() here?  Its used in mhi_queue_buf(), 
which seems to be why the issue doesn't appear there.  It looks like we 
have 3 instances of the same code, which all need to be kept in sync. 
Seems like this bug exists because they were not in sync.

Simplifying the code by removing the duplication seems to be the better 
approach by not only addressing this issue, but also preventing future ones.

>   
>   	/* increment WP */
>   	mhi_add_ring_element(mhi_cntrl, tre_ring);
> @@ -1006,6 +1007,7 @@ int mhi_queue_dma(struct mhi_device *mhi_dev, enum dma_data_direction dir,
>   	struct mhi_ring *buf_ring = &mhi_chan->buf_ring;
>   	struct mhi_buf_info *buf_info;
>   	struct mhi_tre *mhi_tre;
> +	int bei;
>   
>   	/* If MHI host pre-allocates buffers then client drivers cannot queue */
>   	if (mhi_chan->pre_alloc)
> @@ -1043,10 +1045,11 @@ int mhi_queue_dma(struct mhi_device *mhi_dev, enum dma_data_direction dir,
>   	buf_info->len = len;
>   
>   	mhi_tre = tre_ring->wp;
> +	bei = !!(mhi_chan->intmod);
>   
>   	mhi_tre->ptr = MHI_TRE_DATA_PTR(buf_info->p_addr);
>   	mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(buf_info->len);
> -	mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(1, 1, 0, 0);
> +	mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, 1, 0, 0);
>   
>   	/* increment WP */
>   	mhi_add_ring_element(mhi_cntrl, tre_ring);


--
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.