On Thu, Jan 29, 2015 at 12:19:39PM +0000, Zubair Lutfullah Kakakhel wrote:
> From: Alex Smith <alex.smith@xxxxxxxxxx>
>
> This patch adds a driver for the DMA controller found in the Ingenic
> JZ4780.
>
> It currently does not implement any support for the programmable firmware
> feature of the controller - this is not necessary for most uses. It also
> does not take priority into account when allocating channels, it just
> allocates the first available channel. This can be implemented later.
>
> +++ b/drivers/dma/dma-jz4780.c
> @@ -0,0 +1,856 @@
> +/*
> + * Ingenic JZ4780 DMA controller
> + *
> + * Copyright (c) 2013 Imagination Technologies
nitpick we are in '15 now
> +static uint32_t jz4780_dma_transfer_size(unsigned long val, int *ord)
> +{
> + *ord = ffs(val) - 1;
> +
> + /* 8 byte transfer sizes unsupported so fall back on 4. */
> + if (*ord == 3)
> + *ord = 2;
> + else if (*ord > 7)
> + *ord = 7;
> +
> + switch (*ord) {
> + case 0:
> + return JZ_DMA_SIZE_1_BYTE;
> + case 1:
> + return JZ_DMA_SIZE_2_BYTE;
> + case 2:
> + return JZ_DMA_SIZE_4_BYTE;
> + case 4:
> + return JZ_DMA_SIZE_16_BYTE;
> + case 5:
> + return JZ_DMA_SIZE_32_BYTE;
> + case 6:
> + return JZ_DMA_SIZE_64_BYTE;
> + default:
> + return JZ_DMA_SIZE_128_BYTE;
you should fail rather than assuming default for slave, and why aren't we
using the maxburst value for this?
> +
> +static struct dma_async_tx_descriptor *jz4780_dma_prep_dma_cyclic(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_transfer_direction direction,
> + unsigned long flags, void *context)
> +{
> + struct jz4780_dma_chan *jzchan = to_jz4780_dma_chan(chan);
> + struct jz4780_dma_desc *desc;
> + unsigned int periods, i;
> +
> + if (buf_len % period_len)
> + return NULL;
this would be a common requirement, I am wondering if we should add this in
capabilities and let sound-dmaengine layer enforce it
> +static int jz4780_dma_slave_config(struct jz4780_dma_chan *jzchan,
> + const struct dma_slave_config *config)
> +{
> + if ((config->src_addr_width == DMA_SLAVE_BUSWIDTH_8_BYTES)
> + || (config->dst_addr_width == DMA_SLAVE_BUSWIDTH_8_BYTES))
> + return -EINVAL;
> +
> + if ((config->src_addr_width == DMA_SLAVE_BUSWIDTH_UNDEFINED)
> + && (config->dst_addr_width == DMA_SLAVE_BUSWIDTH_UNDEFINED))
> + return -EINVAL;
this is understandable but latter check is not, if this is a limitation pls
mention it here
>
> +static int jz4780_dma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> + unsigned long arg)
> +{
> + struct jz4780_dma_chan *jzchan = to_jz4780_dma_chan(chan);
> + struct dma_slave_config *config = (void *)arg;
> +
> + switch (cmd) {
> + case DMA_TERMINATE_ALL:
> + return jz4780_dma_terminate_all(jzchan);
> + case DMA_SLAVE_CONFIG:
> + return jz4780_dma_slave_config(jzchan, config);
> + default:
> + return -ENOSYS;
> + }
> +}
you haven't kept up with dmaengine development, this needs update.
> +
> +static enum dma_status jz4780_dma_tx_status(struct dma_chan *chan,
> + dma_cookie_t cookie, struct dma_tx_state *txstate)
> +{
> + struct jz4780_dma_chan *jzchan = to_jz4780_dma_chan(chan);
> + struct virt_dma_desc *vdesc;
> + enum dma_status status;
> + unsigned long flags;
> +
> + status = dma_cookie_status(chan, cookie, txstate);
> + if (status == DMA_COMPLETE)
> + return status;
> +
> + spin_lock_irqsave(&jzchan->vchan.lock, flags);
> +
> + vdesc = vchan_find_desc(&jzchan->vchan, cookie);
> + if (vdesc && jzchan->desc && vdesc == &jzchan->desc->vdesc
> + && jzchan->desc->status & (JZ_DMA_DCS_AR | JZ_DMA_DCS_HLT))
> + status = DMA_ERROR;
what about residue reporting?
> +
> +static void jz4780_dma_chan_irq(struct jz4780_dma_dev *jzdma,
> + struct jz4780_dma_chan *jzchan)
> +{
> + uint32_t dcs;
> +
> + spin_lock(&jzchan->vchan.lock);
> +
> + dcs = jz4780_dma_readl(jzdma, JZ_DMA_REG_DCS(jzchan->id));
> + jz4780_dma_writel(jzdma, JZ_DMA_REG_DCS(jzchan->id), 0);
> +
> + if (dcs & JZ_DMA_DCS_AR) {
> + dev_warn(&jzchan->vchan.chan.dev->device,
> + "address error (DCS=0x%x)\n", dcs);
> + }
> +
> + if (dcs & JZ_DMA_DCS_HLT) {
> + dev_warn(&jzchan->vchan.chan.dev->device,
> + "channel halt (DCS=0x%x)\n", dcs);
> + }
> +
> + if (jzchan->desc) {
> + jzchan->desc->status = dcs;
> +
> + if ((dcs & (JZ_DMA_DCS_AR | JZ_DMA_DCS_HLT)) == 0) {
> + if (jzchan->desc->type == DMA_CYCLIC) {
> + vchan_cyclic_callback(&jzchan->desc->vdesc);
> + } else {
> + vchan_cookie_complete(&jzchan->desc->vdesc);
> + jzchan->desc = NULL;
> + }
> +
> + jz4780_dma_begin(jzchan);
this bit is good :)
> +
> + dma_cap_set(DMA_MEMCPY, dd->cap_mask);
> + dma_cap_set(DMA_SLAVE, dd->cap_mask);
> + dma_cap_set(DMA_CYCLIC, dd->cap_mask);
> +
> + dd->dev = dev;
> + dd->chancnt = JZ_DMA_NR_CHANNELS;
> + dd->copy_align = 2; /* 2^2 = 4 byte alignment */
> + dd->device_alloc_chan_resources = jz4780_dma_alloc_chan_resources;
> + dd->device_free_chan_resources = jz4780_dma_free_chan_resources;
> + dd->device_prep_slave_sg = jz4780_dma_prep_slave_sg;
> + dd->device_prep_dma_cyclic = jz4780_dma_prep_dma_cyclic;
> + dd->device_prep_dma_memcpy = jz4780_dma_prep_dma_memcpy;
> + dd->device_control = jz4780_dma_control;
> + dd->device_tx_status = jz4780_dma_tx_status;
> + dd->device_issue_pending = jz4780_dma_issue_pending;
Please initialize controller capabilities as well, this bit is mandatory now
> +
> +static int jz4780_dma_remove(struct platform_device *pdev)
> +{
> + struct jz4780_dma_dev *jzdma = platform_get_drvdata(pdev);
> +
> + of_dma_controller_free(pdev->dev.of_node);
> + dma_async_device_unregister(&jzdma->dma_device);
you are unregistering device even though you can still get an
irq...
--
~Vinod
--
To unsubscribe from this list: send the line "unsubscribe dmaengine" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
