linux-kernel - Re: [PATCH v2 3/3] soc: qcom: geni: Add support for gpi dma

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]

Date:   Fri, 18 Jun 2021 12:23:48 -0500
From:   Bjorn Andersson <bjorn.andersson@...aro.org>
To:     Vinod Koul <vkoul@...nel.org>
Cc:     linux-arm-msm@...r.kernel.org, Andy Gross <agross@...nel.org>,
        Matthias Kaehlcke <mka@...omium.org>,
        Douglas Anderson <dianders@...omium.org>,
        Sumit Semwal <sumit.semwal@...aro.org>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 3/3] soc: qcom: geni: Add support for gpi dma

On Fri 18 Jun 09:18 CDT 2021, Vinod Koul wrote:

> GPI DMA is one of the DMA modes supported on geni, this adds support to
> enable that mode

I think you're missing an opportunity to describe what GPI DMA is.
Perhaps something like:

In GPI DMA mode the serial engine uses the DMA controller found in the
associated wrapper to perform its transaction, add support for enabling
this mode in the engine.

> 
> Signed-off-by: Vinod Koul <vkoul@...nel.org>
> ---
>  drivers/soc/qcom/qcom-geni-se.c | 32 +++++++++++++++++++++++++++++++-
>  include/linux/qcom-geni-se.h    |  5 +++--
>  2 files changed, 34 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/soc/qcom/qcom-geni-se.c b/drivers/soc/qcom/qcom-geni-se.c
> index 08d645b90ed3..40a0a1f88070 100644
> --- a/drivers/soc/qcom/qcom-geni-se.c
> +++ b/drivers/soc/qcom/qcom-geni-se.c
> @@ -307,6 +307,32 @@ static void geni_se_select_dma_mode(struct geni_se *se)
>  		writel_relaxed(val, se->base + SE_GENI_DMA_MODE_EN);
>  }
>  
> +static void geni_se_select_gpi_mode(struct geni_se *se)
> +{
> +	unsigned int gpi_event_en;
> +	unsigned int m_irq_en;
> +	unsigned int s_irq_en;

readl and writel operates on u32, so better to use that.

> +
> +	geni_se_irq_clear(se);
> +	writel(0, se->base + SE_IRQ_EN);
> +
> +	s_irq_en = readl_relaxed(se->base + SE_GENI_S_IRQ_EN);

I don't mind this being _relaxed if done on purpose, but as it's the
only one (and there's no comment) I get the feeling that it's a mistake.

> +	s_irq_en &= ~S_CMD_DONE_EN;
> +	writel(s_irq_en, se->base + SE_GENI_S_IRQ_EN);

The use of 3 different variables (gpi_event_en, m_irq_en, s_irq_en)
forced me to really look if the three sets of operations somehow reused
previous results.

To clarify that this isn't the case I would suggest that you use a
single variable ("val"?) instead.

> +
> +	m_irq_en = readl(se->base + SE_GENI_M_IRQ_EN);
> +	m_irq_en &= ~(M_CMD_DONE_EN | M_TX_FIFO_WATERMARK_EN |
> +		      M_RX_FIFO_WATERMARK_EN | M_RX_FIFO_LAST_EN);
> +	writel(m_irq_en, se->base + SE_GENI_M_IRQ_EN);
> +
> +	writel(GENI_DMA_MODE_EN, se->base + SE_GENI_DMA_MODE_EN);
> +
> +	gpi_event_en = readl(se->base + SE_GSI_EVENT_EN);
> +	gpi_event_en |= (DMA_RX_EVENT_EN | DMA_TX_EVENT_EN |
> +			 GENI_M_EVENT_EN | GENI_S_EVENT_EN);
> +	writel(gpi_event_en, se->base + SE_GSI_EVENT_EN);
> +}
> +
>  /**
>   * geni_se_select_mode() - Select the serial engine transfer mode
>   * @se:		Pointer to the concerned serial engine.
> @@ -314,7 +340,8 @@ static void geni_se_select_dma_mode(struct geni_se *se)
>   */
>  void geni_se_select_mode(struct geni_se *se, enum geni_se_xfer_mode mode)
>  {
> -	WARN_ON(mode != GENI_SE_FIFO && mode != GENI_SE_DMA);
> +	WARN_ON(mode != GENI_SE_FIFO && mode != GENI_SE_DMA &&
> +		mode != GENI_GPI_DMA);

This line can be left unbroken.

>  
>  	switch (mode) {
>  	case GENI_SE_FIFO:
> @@ -323,6 +350,9 @@ void geni_se_select_mode(struct geni_se *se, enum geni_se_xfer_mode mode)
>  	case GENI_SE_DMA:
>  		geni_se_select_dma_mode(se);
>  		break;
> +	case GENI_GPI_DMA:
> +		geni_se_select_gpi_mode(se);
> +		break;
>  	case GENI_SE_INVALID:
>  	default:
>  		break;
> diff --git a/include/linux/qcom-geni-se.h b/include/linux/qcom-geni-se.h
> index 5fda675c5cfe..336b682392b1 100644
> --- a/include/linux/qcom-geni-se.h
> +++ b/include/linux/qcom-geni-se.h
> @@ -11,8 +11,9 @@
>  /* Transfer mode supported by GENI Serial Engines */
>  enum geni_se_xfer_mode {
>  	GENI_SE_INVALID,
> -	GENI_SE_FIFO,
> -	GENI_SE_DMA,

Is the order significant? Is there a reason why SE_DMA ended up last?

> +	GENI_SE_FIFO, /* FIFO mode */
> +	GENI_GPI_DMA, /* GSI aka GPI DMA mode */

If the TLA soup is too soupy, then perhaps we should just document it as
"external DMA" or "Use wrapper's DMA controller"?

And perhaps use kernel-doc instead?

Regards,
Bjorn

> +	GENI_SE_DMA,  /* SE DMA mode */
>  };
>  
>  /* Protocols supported by GENI Serial Engines */
> -- 
> 2.31.1
>