Hi,

On Thu, Jun 24, 2021 at 10:22 PM Vinod Koul [off-list ref] wrote:

+static void geni_se_select_gpi_mode(struct geni_se *se)
+{
+       u32 val;
+
+       geni_se_irq_clear(se);
+
+       writel(0, se->base + SE_IRQ_EN);
+
+       val = readl(se->base + SE_GENI_S_IRQ_EN);
+       val &= ~S_CMD_DONE_EN;
+       writel(val, se->base + SE_GENI_S_IRQ_EN);
+
+       val = readl(se->base + SE_GENI_M_IRQ_EN);
+       val &= ~(M_CMD_DONE_EN | M_TX_FIFO_WATERMARK_EN |
+                M_RX_FIFO_WATERMARK_EN | M_RX_FIFO_LAST_EN);
+       writel(val, se->base + SE_GENI_M_IRQ_EN);
+
+       writel(GENI_DMA_MODE_EN, se->base + SE_GENI_DMA_MODE_EN);
+
+       val = readl(se->base + SE_GSI_EVENT_EN);
+       val |= (DMA_RX_EVENT_EN | DMA_TX_EVENT_EN | GENI_M_EVENT_EN | GENI_S_EVENT_EN);

nit: the above has some extra parenthesis that aren't needed.

I will continue to assert that all of the "set mode" stuff doesn't
really belong here and should be managed by individual drivers [1].
I'll accept that it doesn't have to block forward progress, though I'm
at least a bit disappointed that we asked Qualcomm to do this over 8
months ago and no action was taken. :(

In any case, this looks OK to me:

Reviewed-by: Douglas Anderson <dianders@chromium.org>


[1] https://lore.kernel.org/r/CAD=FV=VWPqswOXJejyXjYT_Yspdu75ELq42cffN87FrpTwPUQg@mail.gmail.com/ (local)