[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3691c453-b2cb-df6f-d029-e82ac98733c7@linaro.org>
Date: Mon, 21 Jan 2019 19:41:56 +0200
From: Georgi Djakov <georgi.djakov@...aro.org>
To: abailon@...libre.com
Cc: linux-pm@...r.kernel.org, devicetree@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Subject: Re: [RFC PATCH 1/2] Add support of imx7ulp to interconnect framework
Hi Alexandre,
Thanks for submitting this patchset!
On 11/17/18 13:19, abailon@...libre.com wrote:
> From: Alexandre Bailon <abailon@...libre.com>
>
> This is a PoC of interconnect driver for the imx7ulp.
> This scales the clock of nic0 and nic1 (the two interconnects).
> In order too solve some issue with clocks (I will give more details
s/too/to/
> later in code), the bus topology described in the driver differ a little
> from the one in the datasheet.
Is there a public datasheet available?
> In addition, the driver manages nic0 and nic1 as an unique
> interconnect. It would have been better to manage them separately but
> it was harder to do it because of the clocks (again).
> Later, I will add third clock for the MMDC endpoint, which is the
> DDR controller. This will be used to scale the DDR in addition of
> interconnect.
>
> Signed-off-by: Alexandre Bailon <abailon@...libre.com.com>
s/.com.com/.com/
> ---
> drivers/interconnect/Kconfig | 1 +
> drivers/interconnect/Makefile | 1 +
> drivers/interconnect/imx/Kconfig | 9 +
> drivers/interconnect/imx/Makefile | 1 +
> drivers/interconnect/imx/imx7ulp.c | 369 +++++++++++++++++++++++++++++
> 5 files changed, 381 insertions(+)
> create mode 100644 drivers/interconnect/imx/Kconfig
> create mode 100644 drivers/interconnect/imx/Makefile
> create mode 100644 drivers/interconnect/imx/imx7ulp.c
>
> diff --git a/drivers/interconnect/Kconfig b/drivers/interconnect/Kconfig
> index 07a8276fa35a..99955906bea8 100644
> --- a/drivers/interconnect/Kconfig
> +++ b/drivers/interconnect/Kconfig
> @@ -11,5 +11,6 @@ menuconfig INTERCONNECT
> if INTERCONNECT
>
> source "drivers/interconnect/qcom/Kconfig"
> +source "drivers/interconnect/imx/Kconfig"
>
> endif
> diff --git a/drivers/interconnect/Makefile b/drivers/interconnect/Makefile
> index 7944cbca0527..4c8814fc2b4b 100644
> --- a/drivers/interconnect/Makefile
> +++ b/drivers/interconnect/Makefile
> @@ -1,3 +1,4 @@
> # SPDX-License-Identifier: GPL-2.0
> obj-$(CONFIG_INTERCONNECT) += core.o
> obj-$(CONFIG_INTERCONNECT_QCOM) += qcom/
> +obj-$(CONFIG_INTERCONNECT_IMX) += imx/
> diff --git a/drivers/interconnect/imx/Kconfig b/drivers/interconnect/imx/Kconfig
> new file mode 100644
> index 000000000000..5eb88eafc1fd
> --- /dev/null
> +++ b/drivers/interconnect/imx/Kconfig
> @@ -0,0 +1,9 @@
> +config INTERCONNECT_IMX
> + bool "IMX Network-on-Chip interconnect drivers"
I was wondering if this can be a module?
> + depends on INTERCONNECT
The above is redundant.
> + depends on ARCH_MXC
It would be nice if we can support COMPILE_TEST.
> + default y
> +
> +config INTERCONNECT_IMX7ULP
> + tristate "i.MX7ULP NIC-301 interconnect driver"
> + depends on INTERCONNECT_IMX
> diff --git a/drivers/interconnect/imx/Makefile b/drivers/interconnect/imx/Makefile
> new file mode 100644
> index 000000000000..61d994b8fe2c
> --- /dev/null
> +++ b/drivers/interconnect/imx/Makefile
> @@ -0,0 +1 @@
> +obj-$(CONFIG_INTERCONNECT_IMX7ULP) += imx7ulp.o
> diff --git a/drivers/interconnect/imx/imx7ulp.c b/drivers/interconnect/imx/imx7ulp.c
> new file mode 100644
> index 000000000000..7715198ad151
> --- /dev/null
> +++ b/drivers/interconnect/imx/imx7ulp.c
> @@ -0,0 +1,369 @@
Missing SPDX?
> +#include <linux/clk.h>
> +#include <linux/device.h>
> +#include <linux/io.h>
> +#include <linux/interconnect-provider.h>
Please move this above io.h to keep it sorted.
> +#include <linux/module.h>
> +#include <linux/of_device.h>
> +#include <linux/of_platform.h>
> +#include <linux/platform_device.h>
> +#include <linux/slab.h>
> +
> +#define IMX7ULP_MAX_LINKS 32
> +
> +/**
> + * Clock property of node
The format should be:
struct my_struct - short description
> + * @name: The name of the clock
> + * @clk: A pointer to the clock
> + * @rate: The rate to use, when the clock frequency is updated
> + */
> +struct imx7ulp_bus_clock {
> + char *name;
> + struct clk *clk;
> + u64 rate;
> +};
> +
> +/**
> + * IMX7ULP specific interconnect nodes
Ditto.
> + * @name: the node name used in debugfs
> + * @id: a unique node identifier
> + * @links: an array of nodes where we can go next while traversing
> + * @num_links: the total number of @links
> + * @buswidth: width (in bit) of the interconnect between a node and the bus
> + * @bus_clk: If not NULL, define clock properties of the node
> + * @ep: true if the node is an end point. Used to scale the frequency
> + */
> +struct imx7ulp_icc_node {
> + char *name;
> + u16 id;
> + u16 links[IMX7ULP_MAX_LINKS];
> + u16 num_links;
> + u16 buswidth;
> + struct imx7ulp_bus_clock *bus_clk;
> + bool ep;
> +};
> +
> +/**
Ditto.
> + * @dev: device pointer for the overall interconnect
> + * Although there are 2 interconnects, they are represented using one
> + * device.
> + * @nodes: Array of nodes
> + * @num_nodes: The total number of nodes
> + * @nic0: Clock properties for nic0 interconnect
> + * @nic1: Clock properties for nic1 interconnect
> + */
> +struct imx7ulp_icc_desc {
> + struct device *dev;
> + struct imx7ulp_icc_node *nodes;
> + size_t num_nodes;
> + struct imx7ulp_bus_clock *nic0;
> + struct imx7ulp_bus_clock *nic1;
> +};
I assume that we can be only change the rate of the interconnect clocks
and there are no other knobs? Is this correct? If that's the case, i
don't see a problem to represent it as a single device.
> +
> +#define DEFINE_AXI_INTERCONNECT(_name, _id, _buswidth, _ep, _clk, \
> + _numlinks, ...) \
> + { \
> + .id = _id, \
> + .name = #_name, \
> + .buswidth = (_buswidth * 4), \
Please put parenthesis around _buswidth or avoid potential precedence
issues.
> + .num_links = _numlinks, \
> + .links = { __VA_ARGS__ }, \
> + .ep = _ep, \
> + .bus_clk = _clk, \
> + }
> +
> +#define DEFINE_AXI_MASTER(_name, _id, _buswidth, dest_id) \
> + DEFINE_AXI_INTERCONNECT(_name, _id, _buswidth, false, NULL, \
> + 1, dest_id)
> +
> +#define DEFINE_AXI_SLAVE(_name, _id, _buswidth, _numlinks, ...) \
> + DEFINE_AXI_INTERCONNECT(_name, _id, _buswidth, false, NULL, \
> + _numlinks, __VA_ARGS__)
> +
> +#define DEFINE_AXI_SLAVE_EP(_name, _id, _buswidth) \
> + DEFINE_AXI_INTERCONNECT(_name, _id, _buswidth, true, NULL, 0, 0)
> +
> +#define DEFINE_AHB_SLAVE_EP(_name, _id) \
> + DEFINE_AXI_SLAVE_EP(_name, _id, 4)
> +
> +static struct imx7ulp_bus_clock nic0_clock = {
> + .name = "nic0_div",
> +};
> +
> +static struct imx7ulp_bus_clock nic1_clock = {
> + .name = "nic1_div",
> +};
> +
> +static struct imx7ulp_icc_node imx7ulp_nodes[] = {
> + /* NIC0 Masters */
> + DEFINE_AXI_MASTER(a7, 2000, 16, 1000),
> + DEFINE_AXI_MASTER(nic1_s1, 2001, 8, 1000),
> + DEFINE_AXI_MASTER(dsi, 2002, 4, 1000),
> + DEFINE_AXI_MASTER(gpu3d, 2003, 8, 1000),
> + DEFINE_AXI_MASTER(gpu2d, 2004, 16, 1000),
> +
> + /* NIC0 Slaves */
> + DEFINE_AXI_SLAVE_EP(mmdc, 2005, 16),
> + DEFINE_AXI_SLAVE_EP(sram0, 2006, 8),
> +
> + /* NIC0 */
> + DEFINE_AXI_INTERCONNECT(nic0_m, 1000, 8, false, &nic0_clock, 1, 1001),
> + DEFINE_AXI_INTERCONNECT(nic0_s, 1001, 8, false, NULL, 3,
> + 2005, 2006, 1008),
> +
> + /* NIC1 Masters */
> + DEFINE_AXI_MASTER(dma1_m, 2011, 8, 1008),
> + DEFINE_AXI_MASTER(caam_m, 2013, 4, 1008),
> + DEFINE_AXI_MASTER(usbotg_m, 2014, 4, 1008),
> + DEFINE_AXI_MASTER(viu_m, 2015, 8, 1008),
> + DEFINE_AXI_MASTER(usdhc0_m, 2016, 4, 1008),
> + DEFINE_AXI_MASTER(usdhc1_m, 2017, 4, 1008),
> +
> + /* NIC1 Slaves */
> + DEFINE_AXI_SLAVE_EP(sram1, 2018, 8),
> + DEFINE_AXI_SLAVE_EP(secure_ram, 2020, 4),
> + DEFINE_AXI_SLAVE_EP(rom, 2023, 4), /* Also flexbus, gpu 2d and 3d */
> + DEFINE_AXI_SLAVE_EP(m4xbar, 2024, 4),
> +
> + /* AHB0 */
> + DEFINE_AHB_SLAVE_EP(dma1, 8),
> + DEFINE_AHB_SLAVE_EP(dmadesc1, 9),
> + DEFINE_AHB_SLAVE_EP(rgpio2p, 15),
> + DEFINE_AHB_SLAVE_EP(flexbus, 16),
> + DEFINE_AHB_SLAVE_EP(sema42_1, 27),
> + DEFINE_AHB_SLAVE_EP(dmamux1, 33),
> + DEFINE_AHB_SLAVE_EP(mu_b, 34),
> + DEFINE_AHB_SLAVE_EP(caam, 36),
> + DEFINE_AHB_SLAVE_EP(tpm4, 37),
> + DEFINE_AHB_SLAVE_EP(tpm5, 38),
> + DEFINE_AHB_SLAVE_EP(lpit1, 39),
> + DEFINE_AHB_SLAVE_EP(lpspi2, 41),
> + DEFINE_AHB_SLAVE_EP(lpspi3, 42),
> + DEFINE_AHB_SLAVE_EP(lpi2c4, 43),
> + DEFINE_AHB_SLAVE_EP(lpi2c5, 44),
> + DEFINE_AHB_SLAVE_EP(lpuart4, 45),
> + DEFINE_AHB_SLAVE_EP(lpuart5, 46),
> + DEFINE_AHB_SLAVE_EP(flexio1, 49),
> + DEFINE_AHB_SLAVE_EP(usbotg1, 51),
> + DEFINE_AHB_SLAVE_EP(usbotg2, 52),
> + DEFINE_AHB_SLAVE_EP(usbphy, 53),
> + DEFINE_AHB_SLAVE_EP(usbpl301, 54),
> + DEFINE_AHB_SLAVE_EP(usdhc0, 55),
> + DEFINE_AHB_SLAVE_EP(usdhc1, 56),
> + DEFINE_AHB_SLAVE_EP(trgmux1, 59),
> + DEFINE_AHB_SLAVE_EP(wdog1, 61),
> + DEFINE_AHB_SLAVE_EP(scg1, 62),
> + DEFINE_AHB_SLAVE_EP(pcc2, 63),
> + DEFINE_AHB_SLAVE_EP(pmc1, 64),
> + DEFINE_AHB_SLAVE_EP(cmc1, 65),
> + DEFINE_AHB_SLAVE_EP(wdog2, 67),
> +
> + DEFINE_AXI_SLAVE(ahb0, 2021, 4, 32,
> + 8, 9, 15, 16, 27, 33, 34, 35, 36, 37, 38, 39, 41, 42,
> + 43, 44, 45, 46, 49, 51, 52, 53, 54, 55, 56, 59, 61, 62,
> + 63, 64, 65, 67),
> +
> + /* AHB1 */
> + DEFINE_AHB_SLAVE_EP(romc1, 116),
> + DEFINE_AHB_SLAVE_EP(tpm6, 133),
> + DEFINE_AHB_SLAVE_EP(tpm7, 134),
> + DEFINE_AHB_SLAVE_EP(lpi2c6, 136),
> + DEFINE_AHB_SLAVE_EP(lpi2c7, 137),
> + DEFINE_AHB_SLAVE_EP(lpuart6, 138),
> + DEFINE_AHB_SLAVE_EP(lpuart7, 139),
> + DEFINE_AHB_SLAVE_EP(viu, 140),
> + DEFINE_AHB_SLAVE_EP(dsi, 141),
> + DEFINE_AHB_SLAVE_EP(lcdif, 142),
> + DEFINE_AHB_SLAVE_EP(mmdc, 143),
> + DEFINE_AHB_SLAVE_EP(iomuxc1, 144),
> + DEFINE_AHB_SLAVE_EP(iomuxc_ddr, 145),
> + DEFINE_AHB_SLAVE_EP(pctlc, 146),
> + DEFINE_AHB_SLAVE_EP(pctld, 147),
> + DEFINE_AHB_SLAVE_EP(pctle, 148),
> + DEFINE_AHB_SLAVE_EP(pctlf, 149),
> + DEFINE_AHB_SLAVE_EP(pcc3, 151),
> +
> + DEFINE_AXI_SLAVE(ahb1, 2022, 4, 19,
> + 116, 133, 134, 136, 137, 138, 139, 140, 141,
> + 142, 143, 144, 145, 146, 147, 148, 149, 151),
> +
> + /* NIC1 */
> + DEFINE_AXI_INTERCONNECT(nic1_m, 1008, 8, false, &nic1_clock, 1, 1009),
> + DEFINE_AXI_INTERCONNECT(nic1_s, 1009, 8, false, NULL, 7,
> + 2018, 1000, 2020, 2021, 2022, 2023, 2024),> +};
> +
> +static struct imx7ulp_icc_desc icc_desc = {
> + .nodes = (struct imx7ulp_icc_node *)&imx7ulp_nodes,
> + .num_nodes = ARRAY_SIZE(imx7ulp_nodes),
> +};
> +
> +static int imx7ulp_icc_aggregate(struct icc_node *node, u32 avg_bw,
> + u32 peak_bw, u32 *agg_avg, u32 *agg_peak)
> +{
> + *agg_avg += avg_bw;
> + *agg_peak = max(*agg_peak, peak_bw);
> +
> + return 0;
> +}
> +
> +static int imx7ulp_icc_set(struct icc_node *src, struct icc_node *dst)
> +{
> + int ret;
> + struct imx7ulp_icc_node *node;
> + struct imx7ulp_icc_desc *desc = src->provider->data;
> +
> + /* If node has a clock, compute the new rate */
> + node = src->data;
> + if (node->bus_clk) {
> + u64 rate;
> + struct imx7ulp_bus_clock *bus_clk = node->bus_clk;
> +
> + rate = (src->peak_bw / node->buswidth) * 1000;
Maybe use do_div().
> + rate = clk_round_rate(bus_clk->clk, rate);
> + if (rate != bus_clk->rate) {
> + dev_dbg(desc->dev, "%s: new rate = %llu\n",
> + bus_clk->name, rate);
> + bus_clk->rate = rate;
> + }
> + }
> +
> + /*
> + * If the node is an endpoint, update the clock rate of nic0 and nic1
> + * nodes. nic1 clock derives from nic0, so we usually have to update
> + * both of them even if bandwidth changed only for one of the
> + * interconnect.
> + */
> + node = dst->data;
> + if (node->ep) {
> + struct imx7ulp_bus_clock *nic0_clk = desc->nic0;
> + struct imx7ulp_bus_clock *nic1_clk = desc->nic1;
> +
> + if (nic1_clk->rate > nic0_clk->rate)
> + nic0_clk->rate = nic1_clk->rate;
> +
> + if (clk_get_rate(nic0_clk->clk) != nic0_clk->rate) {
> + dev_dbg(desc->dev, "%s: apply rate = %llu\n",
> + nic0_clk->name, nic0_clk->rate);
> + ret = clk_set_rate(nic0_clk->clk, nic0_clk->rate);
> + if (ret)
> + dev_err(desc->dev, "Failed to set clock: %d\n",
> + ret);
> + }
> +
> + if (clk_get_rate(nic1_clk->clk) != nic1_clk->rate) {
> + dev_dbg(desc->dev, "%s: apply rate = %llu\n",
> + nic1_clk->name, nic1_clk->rate);
> + ret = clk_set_rate(nic1_clk->clk, nic1_clk->rate);
> + if (ret)
> + dev_err(desc->dev, "Failed to set clock: %d\n",
> + ret);
> + }
> + }
> +
> + return 0;
Just f.y.i, i recently introduced another patch that reverts to the
settings to the previous configuration if any of the requests fail, so
you can also return errors here (if it makes sense to do so on this
platform).
> +}
> +
> +static int imx7ulp_probe(struct platform_device *pdev)
> +{
> + struct imx7ulp_icc_desc *desc = &icc_desc;
> + struct icc_provider *provider;
> + size_t i;
> + int ret;
> +
> + desc->dev = &pdev->dev;
> + desc->nic0 = &nic0_clock;
> + desc->nic1 = &nic1_clock;
> +
> + provider = devm_kzalloc(&pdev->dev, sizeof(*provider), GFP_KERNEL);
> + provider->set = imx7ulp_icc_set;
> + provider->aggregate = imx7ulp_icc_aggregate;
If you want to use DT, you will need to put the endpoint IDs in some
header file and implement the xlate() function or use the generic
of_icc_xlate_onecell().
> + INIT_LIST_HEAD(&provider->nodes);
> + provider->data = desc;
> +
> + ret = icc_provider_add(provider);
> + if (ret) {
> + dev_err(&pdev->dev, "error adding interconnect provider\n");
> + return ret;
> + }
> +
> + for (i = 0; i < desc->num_nodes; i++) {
> + struct icc_node *node;
> + int ret;
> + size_t j;
> +
> + if (desc->nodes[i].bus_clk) {
> + struct imx7ulp_bus_clock *bus_clk;
> +
> + bus_clk = desc->nodes[i].bus_clk;
> + bus_clk->clk = devm_clk_get(&pdev->dev, bus_clk->name);
> + if (IS_ERR(bus_clk->clk)) {
> + dev_err(&pdev->dev,
> + "Failed to get clock %s: %d\n",
> + bus_clk->name, PTR_ERR(bus_clk->clk));
> + ret = PTR_ERR(bus_clk->clk);
> + goto err;
> + }
> +
> + ret = clk_prepare_enable(bus_clk->clk);
> + if (ret) {
> + dev_err(&pdev->dev,
> + "error enabling clock: %s %d\n",
> + bus_clk->name, ret);
> + goto err;
> + }
> + }
> +
> + node = icc_node_create(desc->nodes[i].id);
> + if (IS_ERR(node)) {
> + ret = PTR_ERR(node);
> + dev_err(&pdev->dev, "Failed to create node %s\n",
> + desc->nodes[i].name);
> + goto err;
> + }
> +
> + node->name = desc->nodes[i].name;
> + node->data = &desc->nodes[i];
> + icc_node_add(node, provider);
> +
> + dev_dbg(&pdev->dev, "registered node %p %s %d\n", node,
> + desc->nodes[i].name, node->id);
> +
> + for (j = 0; j < desc->nodes[i].num_links; j++)
> + if (desc->nodes[i].links[j])
> + icc_link_create(node, desc->nodes[i].links[j]);
> + }
> +
> + platform_set_drvdata(pdev, provider);
> +
> + return 0;
> +err:
> + icc_provider_del(provider);
First delete the nodes (if any) and then the provider.
> + return ret;
> +}
> +
> +static int imx7ulp_remove(struct platform_device *pdev)
> +{
> + struct icc_provider *provider = platform_get_drvdata(pdev);
> +
> + icc_provider_del(provider);
Ditto.
> + return 0;
> +}
> +
> +static const struct of_device_id imx7ulp_of_match[] = {
> + { .compatible = "fsl,imx7ulp-icc" },
> + { },
> +};
> +MODULE_DEVICE_TABLE(of, imx7ulp_of_match);
> +
> +static struct platform_driver imx7ulp_driver = {
> + .probe = imx7ulp_probe,
> + .remove = imx7ulp_remove,
> + .driver = {
> + .name = "imx7ulp-icc",
> + .of_match_table = imx7ulp_of_match,
> + },
> +};
> +module_platform_driver(imx7ulp_driver);
Please use builtin_platform_driver if it can't be a module.
Thank you for working on this! I am expecting the next version.
BR,
Georgi
Powered by blists - more mailing lists