Den 08.03.2022 02.56, skrev Yunhao Tian:

In __spi_validate, there's a validation that no partial transfers are accepted (xfer->len % w_size must be zero). When max_chunk is not a multiple of bpw (e.g.max_chunk = 65535, bpw = 16), the transfer will be rejected.

This patch clamps max_chunk to the word size, preventing

I think align is a better word here than clamp.

the transfer from being rejected.

Signed-off-by: Yunhao Tian t123yh.xyz@gmail.com

drivers/gpu/drm/drm_mipi_dbi.c | 9 +++++++++ 1 file changed, 9 insertions(+)

diff --git a/drivers/gpu/drm/drm_mipi_dbi.c b/drivers/gpu/drm/drm_mipi_dbi.c index 71b646c4131f..440dc9fec6cc 100644 --- a/drivers/gpu/drm/drm_mipi_dbi.c +++ b/drivers/gpu/drm/drm_mipi_dbi.c @@ -1182,6 +1182,15 @@ int mipi_dbi_spi_transfer(struct spi_device *spi, u32 speed_hz, struct spi_message m; size_t chunk; int ret;

• int w_size;
• if (bpw <= 8)

• w_size = 1;
  

• else if (bpw <= 16)

• w_size = 2;
  

• else

• w_size = 4;
  

• max_chunk -= (max_chunk % w_size);

mipi_dbi_spi_transfer() is only called with bpw= 8 or 16, so I think this can be simplified to: max_chunk = ALIGN_DOWN(max_chunk, 2); We might shorten the max transfer by one byte when bpw=8, but that doesn't matter. A short comment explaining why we need this would be nice.

Please add a fixes tag so the patch is backported to the stable kernels:

Fixes: d23d4d4dac01 ("drm/tinydrm: Move tinydrm_spi_transfer()")

Noralf.

spi_message_init_with_transfers(&m, &tr, 1);