[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <d641c361-278d-4ebe-88be-c872290e474f@linaro.org>
Date: Mon, 9 Dec 2024 15:35:23 +0100
From: neil.armstrong@...aro.org
To: Jyothi Kumar Seerapu <quic_jseerapu@...cinc.com>,
Vinod Koul <vkoul@...nel.org>
Cc: linux-arm-msm@...r.kernel.org, dmaengine@...r.kernel.org,
linux-kernel@...r.kernel.org, quic_msavaliy@...cinc.com,
quic_vtanuku@...cinc.com
Subject: Re: [PATCH v5] dmaengine: qcom: gpi: Add GPI immediate DMA support
for SPI protocol
On 09/12/2024 08:50, Jyothi Kumar Seerapu wrote:
> The DMA TRE(Transfer ring element) buffer contains the DMA
> buffer address. Accessing data from this address can cause
> significant delays in SPI transfers, which can be mitigated to
> some extent by utilizing immediate DMA support.
>
> QCOM GPI DMA hardware supports an immediate DMA feature for data
> up to 8 bytes, storing the data directly in the DMA TRE buffer
> instead of the DMA buffer address. This enhancement enables faster
> SPI data transfers.
>
> This optimization reduces the average transfer time from 25 us to
> 16 us for a single SPI transfer of 8 bytes length, with a clock
> frequency of 50 MHz.
>
> Signed-off-by: Jyothi Kumar Seerapu <quic_jseerapu@...cinc.com>
> ---
> v4 -> v5:
> - For Immediate DMA, instead of making dma type as 0x10 and then
> enable 16th bit of dword3, directly updating the dma type as 0x11.
>
> Link to v4:
> https://lore.kernel.org/all/20241205170611.18566-1-quic_jseerapu@quicinc.com/
>
> v3 -> v4:
> - Instead using extra variable(immediate_dma) for Immediate dma
> condition check, made it to inlined.
> - Removed the extra brackets around Immediate dma condition check.
>
> Link to v3:
> https://lore.kernel.org/lkml/20241204122059.24239-1-quic_jseerapu@quicinc.com/
>
> v2 -> v3:
> - When to enable Immediate DMA support, control is moved to GPI driver
> from SPI driver.
> - Optimizations are done in GPI driver related to immediate dma changes.
> - Removed the immediate dma supported changes in qcom-gpi-dma.h file
> and handled in GPI driver.
>
> Link to v2:
> https://lore.kernel.org/all/20241128133351.24593-2-quic_jseerapu@quicinc.com/
> https://lore.kernel.org/all/20241128133351.24593-3-quic_jseerapu@quicinc.com/
>
> v1 -> v2:
> - Separated the patches to dmaengine and spi subsystems
> - Removed the changes which are not required for this feature from
> qcom-gpi-dma.h file.
> - Removed the type conversions used in gpi_create_spi_tre.
>
> Link to v1:
> https://lore.kernel.org/lkml/20241121115201.2191-2-quic_jseerapu@quicinc.com/
>
> drivers/dma/qcom/gpi.c | 31 +++++++++++++++++++++++++------
> 1 file changed, 25 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/dma/qcom/gpi.c b/drivers/dma/qcom/gpi.c
> index 52a7c8f2498f..b1f0001cc99c 100644
> --- a/drivers/dma/qcom/gpi.c
> +++ b/drivers/dma/qcom/gpi.c
> @@ -18,6 +18,7 @@
> #include "../virt-dma.h"
>
> #define TRE_TYPE_DMA 0x10
> +#define TRE_TYPE_IMMEDIATE_DMA 0x11
> #define TRE_TYPE_GO 0x20
> #define TRE_TYPE_CONFIG0 0x22
>
> @@ -64,6 +65,7 @@
>
> /* DMA TRE */
> #define TRE_DMA_LEN GENMASK(23, 0)
> +#define TRE_DMA_IMMEDIATE_LEN GENMASK(3, 0)
>
> /* Register offsets from gpi-top */
> #define GPII_n_CH_k_CNTXT_0_OFFS(n, k) (0x20000 + (0x4000 * (n)) + (0x80 * (k)))
> @@ -1711,6 +1713,7 @@ static int gpi_create_spi_tre(struct gchan *chan, struct gpi_desc *desc,
> dma_addr_t address;
> struct gpi_tre *tre;
> unsigned int i;
> + int len;
>
> /* first create config tre if applicable */
> if (direction == DMA_MEM_TO_DEV && spi->set_config) {
> @@ -1763,14 +1766,30 @@ static int gpi_create_spi_tre(struct gchan *chan, struct gpi_desc *desc,
> tre_idx++;
>
> address = sg_dma_address(sgl);
> - tre->dword[0] = lower_32_bits(address);
> - tre->dword[1] = upper_32_bits(address);
> + len = sg_dma_len(sgl);
>
> - tre->dword[2] = u32_encode_bits(sg_dma_len(sgl), TRE_DMA_LEN);
> + /* Support Immediate dma for write transfers for data length up to 8 bytes */
> + if (direction == DMA_MEM_TO_DEV && len <= 2 * sizeof(tre->dword[0])) {
> + /*
> + * For Immediate dma, data length may not always be length of 8 bytes,
> + * it can be length less than 8, hence initialize both dword's with 0
> + */
> + tre->dword[0] = 0;
> + tre->dword[1] = 0;
> + memcpy(&tre->dword[0], sg_virt(sgl), len);
>
> - tre->dword[3] = u32_encode_bits(TRE_TYPE_DMA, TRE_FLAGS_TYPE);
> - if (direction == DMA_MEM_TO_DEV)
> - tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_IEOT);
> + tre->dword[2] = u32_encode_bits(len, TRE_DMA_IMMEDIATE_LEN);
> + tre->dword[3] = u32_encode_bits(TRE_TYPE_IMMEDIATE_DMA, TRE_FLAGS_TYPE);
> + } else {
> + tre->dword[0] = lower_32_bits(address);
> + tre->dword[1] = upper_32_bits(address);
> +
> + tre->dword[2] = u32_encode_bits(len, TRE_DMA_LEN);
> + tre->dword[3] = u32_encode_bits(TRE_TYPE_DMA, TRE_FLAGS_TYPE);
> + }
> +
> + tre->dword[3] |= u32_encode_bits(direction == DMA_MEM_TO_DEV,
> + TRE_FLAGS_IEOT);
>
> for (i = 0; i < tre_idx; i++)
> dev_dbg(dev, "TRE:%d %x:%x:%x:%x\n", i, desc->tre[i].dword[0],
Reviewed-by: Neil Armstrong <neil.armstrong@...aro.org>
Powered by blists - more mailing lists