On Thu, May 10, 2018 at 01:59:37PM +0530, Rajesh Yadav wrote:
> SoCs containing dpu have a MDSS top level wrapper
> which includes sub-blocks as dpu, dsi, phy, dp etc.
> MDSS top level wrapper manages common resources like
> common clocks, power and irq for its sub-blocks.
>
> Currently, in dpu driver, all the power resource
> management is part of power_handle which manages
> these resources via a custom implementation. And
> the resource relationships are not modelled properly
> in dt. Moreover the irq domain handling code is part
> of dpu device (which is a child device) due to lack
> of a dedicated driver for MDSS top level wrapper
> device.
>
> This change adds dpu_mdss top level driver to handle
> common clock like - core clock, ahb clock
> (for register access), main power supply (i.e. gdsc)
> and irq management.
> The top level mdss device/driver acts as an interrupt
> controller and manage hwirq mapping for its child
> devices.
>
> It implements runtime_pm support for resource management.
> Child nodes can control these resources via runtime_pm
> get/put calls on their corresponding devices due to parent
> child relationship defined in dt.
>
> Signed-off-by: Rajesh Yadav <ryadav@xxxxxxxxxxxxxx>
> ---
> drivers/gpu/drm/msm/Makefile | 1 +
> drivers/gpu/drm/msm/disp/dpu1/dpu_core_irq.c | 97 -------
> drivers/gpu/drm/msm/disp/dpu1/dpu_core_irq.h | 14 -
> drivers/gpu/drm/msm/disp/dpu1/dpu_hw_catalog.c | 9 -
> drivers/gpu/drm/msm/disp/dpu1/dpu_hw_catalog.h | 7 -
> drivers/gpu/drm/msm/disp/dpu1/dpu_hw_interrupts.c | 29 +--
> drivers/gpu/drm/msm/disp/dpu1/dpu_hw_interrupts.h | 11 -
> drivers/gpu/drm/msm/disp/dpu1/dpu_irq.c | 48 +---
> drivers/gpu/drm/msm/disp/dpu1/dpu_kms.c | 6 -
> drivers/gpu/drm/msm/disp/dpu1/dpu_kms.h | 2 -
> drivers/gpu/drm/msm/disp/dpu1/dpu_mdss.c | 301 ++++++++++++++++++++++
> drivers/gpu/drm/msm/dpu_io_util.c | 55 ++++
> drivers/gpu/drm/msm/msm_drv.c | 26 +-
> drivers/gpu/drm/msm/msm_drv.h | 2 +-
> drivers/gpu/drm/msm/msm_kms.h | 2 +
> include/linux/dpu_io_util.h | 2 +
> 16 files changed, 390 insertions(+), 222 deletions(-)
> create mode 100644 drivers/gpu/drm/msm/disp/dpu1/dpu_mdss.c
<snip>
> -static struct dpu_mdss_base_cfg *__intr_offset(struct dpu_mdss_cfg *m,
> +static int __intr_offset(struct dpu_mdss_cfg *m,
> void __iomem *addr, struct dpu_hw_blk_reg_map *hw)
> {
> if (!m || !addr || !hw || m->mdp_count == 0)
> - return NULL;
> + return -EINVAL;
It looks like this is only used by dpu_hw_intr_init and you know that m, addr
and hw will all be valid and I'm assuming that if we get this far there is at
least some assumption that mdp_count is set so it looks like you can skip this
error path and make this a void function.
> hw->base_off = addr;
> - hw->blk_off = m->mdss[0].base;
> + hw->blk_off = m->mdp[0].base;
> hw->hwversion = m->hwversion;
> - return &m->mdss[0];
> + return 0;
> }
>
> struct dpu_hw_intr *dpu_hw_intr_init(void __iomem *addr,
> struct dpu_mdss_cfg *m)
> {
> struct dpu_hw_intr *intr;
> - struct dpu_mdss_base_cfg *cfg;
> + int ret;
>
> if (!addr || !m)
> return ERR_PTR(-EINVAL);
> @@ -1195,10 +1182,10 @@ struct dpu_hw_intr *dpu_hw_intr_init(void __iomem *addr,
> if (!intr)
> return ERR_PTR(-ENOMEM);
>
> - cfg = __intr_offset(m, addr, &intr->hw);
> - if (!cfg) {
> + ret = __intr_offset(m, addr, &intr->hw);
> + if (ret) {
> kfree(intr);
> - return ERR_PTR(-EINVAL);
> + return ERR_PTR(ret);
> }
And making that a void function has the added benefit of skipping all of this
> __setup_intr_ops(&intr->ops);
>
<snip>
> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_mdss.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_mdss.c
> new file mode 100644
> index 0000000..26bf2c1
> --- /dev/null
> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_mdss.c
> @@ -0,0 +1,301 @@
> +/*
> + * SPDX-License-Identifier: GPL-2.0
> + * Copyright (c) 2018, The Linux Foundation
> + */
> +
> +#include "dpu_kms.h"
> +
> +#define to_dpu_mdss(x) container_of(x, struct dpu_mdss, base)
> +
> +#define HW_INTR_STATUS 0x0010
> +
> +struct dpu_mdss {
> + struct msm_mdss base;
> + void __iomem *mmio;
> + unsigned long mmio_len;
> + u32 hwversion;
> + struct dss_module_power mp;
> + struct dpu_irq_controller irq_controller;
> +};
> +
> +static inline void _dpu_mdss_hw_rev_init(struct dpu_mdss *dpu_mdss)
> +{
> + dpu_mdss->hwversion = readl_relaxed(dpu_mdss->mmio + 0x0);
> +}
> +
> +static int _dpu_mdss_get_intr_sources(struct dpu_mdss *dpu_mdss,
> + uint32_t *sources)
> +{
> + *sources = readl_relaxed(dpu_mdss->mmio + HW_INTR_STATUS);
> + return 0;
> +}
It seems like this could just as easily be:
static u32 _dpu_mdss_get_intr_sources(struct dpu_mdss *dpu_mdss)
{
return readl_relaxed(dpu_mdss->mmio + HW_INTR_STATUS);
}
Or it shouldn't return a 0 for no great reason but if you are going to go
out of your way to create a sub function I feel like it should do work.
> +static irqreturn_t dpu_mdss_irq(int irq, void *arg)
> +{
> + struct dpu_mdss *dpu_mdss = arg;
> + u32 interrupts;
> +
> + if (!arg)
> + return IRQ_NONE;
This is only possible if you are passing a NULL argument to the handler which I
doubt you are.
> + _dpu_mdss_get_intr_sources(dpu_mdss, &interrupts);
> +
See how this could just as easily be
interrupts = _dpu_mdss_get_intr_sources(dpu_mdss);
> + while (interrupts) {
> + irq_hw_number_t hwirq = fls(interrupts) - 1;
> + unsigned int mapping;
> + int rc;
> +
> + mapping = irq_find_mapping(dpu_mdss->irq_controller.domain,
> + hwirq);
> + if (mapping == 0) {
> + DPU_EVT32(hwirq, DPU_EVTLOG_ERROR);
> + goto error;
> + }
> +
> + rc = generic_handle_irq(mapping);
> + if (rc < 0) {
> + DPU_EVT32(hwirq, mapping, rc, DPU_EVTLOG_ERROR);
> + goto error;
> + }
> +
> + interrupts &= ~(1 << hwirq);
> + }
> +
> + return IRQ_HANDLED;
> +
> +error:
> + /* bad situation, inform irq system, it may disable overall MDSS irq */
> + return IRQ_NONE;
> +}
> +
> +static void dpu_mdss_irq_mask(struct irq_data *irqd)
> +{
> + struct dpu_mdss *dpu_mdss;
> +
> + if (!irqd || !irq_data_get_irq_chip_data(irqd)) {
> + DPU_ERROR("invalid parameters irqd %d\n", irqd != NULL);
> + return;
> + }
Neither of these can be NULL at this point since you specify the chip data
from the domain host_data which is known good. You don't need these checks.
> + dpu_mdss = irq_data_get_irq_chip_data(irqd);
> +
> + /* memory barrier */
> + smp_mb__before_atomic();
> + clear_bit(irqd->hwirq, &dpu_mdss->irq_controller.enabled_mask);
> + /* memory barrier */
> + smp_mb__after_atomic();
> +}
> +
> +static void dpu_mdss_irq_unmask(struct irq_data *irqd)
> +{
> + struct dpu_mdss *dpu_mdss;
> +
> + if (!irqd || !irq_data_get_irq_chip_data(irqd)) {
> + DPU_ERROR("invalid parameters irqd %d\n", irqd != NULL);
> + return;
> + }
Same as above.
> + dpu_mdss = irq_data_get_irq_chip_data(irqd);
> +
> + /* memory barrier */
> + smp_mb__before_atomic();
> + set_bit(irqd->hwirq, &dpu_mdss->irq_controller.enabled_mask);
> + /* memory barrier */
> + smp_mb__after_atomic();
> +}
> +
> +static struct irq_chip dpu_mdss_irq_chip = {
> + .name = "dpu_mdss",
> + .irq_mask = dpu_mdss_irq_mask,
> + .irq_unmask = dpu_mdss_irq_unmask,
> +};
> +
> +static int dpu_mdss_irqdomain_map(struct irq_domain *domain,
> + unsigned int irq, irq_hw_number_t hwirq)
> +{
> + struct dpu_mdss *dpu_mdss;
> + int ret;
> +
> + if (!domain || !domain->host_data) {
> + DPU_ERROR("invalid parameters domain %d\n", domain != NULL);
> + return -EINVAL;
> + }
The pointer to this function is derived from the domain pointer so it is known
good. You specify host_data when allocating the domain from a known good
pointer. Neither of these checks are needed.
> + dpu_mdss = domain->host_data;
> +
> + irq_set_chip_and_handler(irq, &dpu_mdss_irq_chip, handle_level_irq);
> + ret = irq_set_chip_data(irq, dpu_mdss);
> +
> + return ret;
> +}
> +
> +static const struct irq_domain_ops dpu_mdss_irqdomain_ops = {
> + .map = dpu_mdss_irqdomain_map,
> + .xlate = irq_domain_xlate_onecell,
> +};
> +
> +static int _dpu_mdss_irq_domain_add(struct dpu_mdss *dpu_mdss)
> +{
> + struct device *dev;
> + struct irq_domain *domain;
> +
> + dev = dpu_mdss->base.dev->dev;
> +
> + domain = irq_domain_add_linear(dev->of_node, 32,
> + &dpu_mdss_irqdomain_ops, dpu_mdss);
> + if (!domain) {
> + DPU_ERROR("failed to add irq_domain\n");
> + return -EINVAL;
> + }
> +
> + dpu_mdss->irq_controller.enabled_mask = 0;
> + dpu_mdss->irq_controller.domain = domain;
> +
> + return 0;
> +}
> +
> +int _dpu_mdss_irq_domain_fini(struct dpu_mdss *dpu_mdss)
> +{
> + if (dpu_mdss->irq_controller.domain) {
> + irq_domain_remove(dpu_mdss->irq_controller.domain);
> + dpu_mdss->irq_controller.domain = NULL;
> + }
> + return 0;
> +}
> +static int dpu_mdss_enable(struct msm_mdss *mdss)
> +{
> + struct dpu_mdss *dpu_mdss = to_dpu_mdss(mdss);
> + struct dss_module_power *mp = &dpu_mdss->mp;
> + int ret = 0;
You don't need to initialize ret.
> + ret = msm_dss_enable_clk(mp->clk_config, mp->num_clk, true);
> + if (ret)
> + DPU_ERROR("clock enable failed, ret:%d\n", ret);
> +
> + return 0;
You should return 'ret' here - presumably the calling function cares about
the result.
> +}
> +
> +static int dpu_mdss_disable(struct msm_mdss *mdss)
> +{
> + struct dpu_mdss *dpu_mdss = to_dpu_mdss(mdss);
> + struct dss_module_power *mp = &dpu_mdss->mp;
> + int ret = 0;
You don't need to initialize ret.
> + ret = msm_dss_enable_clk(mp->clk_config, mp->num_clk, false);
> + if (ret)
> + DPU_ERROR("clock disable failed, ret:%d\n", ret);
> +
> + return 0;
You should return 'ret' to the calling function.
> +}
> +
> +static void dpu_mdss_destroy(struct drm_device *dev)
> +{
> + struct platform_device *pdev = to_platform_device(dev->dev);
> + struct msm_drm_private *priv = dev->dev_private;
> + struct dpu_mdss *dpu_mdss = to_dpu_mdss(priv->mdss);
> + struct dss_module_power *mp = &dpu_mdss->mp;
> +
> + if (!dpu_mdss)
> + return;
to_dpu_mdss is a container_of() so this can never be NULL. The check should
be on priv->mdss instead.
> + _dpu_mdss_irq_domain_fini(dpu_mdss);
> +
> + msm_dss_put_clk(mp->clk_config, mp->num_clk);
> + devm_kfree(&pdev->dev, mp->clk_config);
You don't need to free these if you went out of your way to use devm in the
first place.
> +
> + if (dpu_mdss->mmio)
> + msm_iounmap(pdev, dpu_mdss->mmio);
> + dpu_mdss->mmio = NULL;
> +
> + pm_runtime_disable(dev->dev);
> +
> + devm_kfree(dev->dev, dpu_mdss);
> + priv->mdss = NULL;
> +}
> +
> +static const struct msm_mdss_funcs mdss_funcs = {
> + .enable = dpu_mdss_enable,
> + .disable = dpu_mdss_disable,
> + .destroy = dpu_mdss_destroy,
> +};
> +
> +int dpu_mdss_init(struct drm_device *dev)
> +{
> + struct platform_device *pdev = to_platform_device(dev->dev);
> + struct msm_drm_private *priv = dev->dev_private;
> + struct dpu_mdss *dpu_mdss;
> + struct dss_module_power *mp;
> + int ret = 0;
> +
> + if (!of_device_is_compatible(dev->dev->of_node, "qcom,dpu-mdss"))
> + return 0;
> +
> + dpu_mdss = devm_kzalloc(dev->dev, sizeof(*dpu_mdss), GFP_KERNEL);
> + if (!dpu_mdss)
> + return -ENOMEM;
> +
> + dpu_mdss->mmio = msm_ioremap(pdev, "mdss_phys", "mdss_phys");
> + if (IS_ERR(dpu_mdss->mmio)) {
> + ret = PTR_ERR(dpu_mdss->mmio);
> + DPU_ERROR("mdss register memory map failed: %d\n", ret);
> + dpu_mdss->mmio = NULL;
You don't need to set this to NULL - you are going to immediately free
this structure.
> + goto error;
> + }
> + DRM_INFO("mapped mdss address space @%p\n", dpu_mdss->mmio);
This should be debug or removed. In any event, don't be using %p.
> + dpu_mdss->mmio_len = msm_iomap_size(pdev, "mdss_phys");
> +
> + mp = &dpu_mdss->mp;
> + ret = msm_dss_parse_clock(pdev, mp);
> + if (ret) {
> + DPU_ERROR("failed to parse clocks, ret=%d\n", ret);
> + goto clk_parse_err;
> + }
> +
> + ret = msm_dss_get_clk(&pdev->dev, mp->clk_config, mp->num_clk);
> + if (ret) {
> + DPU_ERROR("failed to get clocks, ret=%d\n", ret);
> + goto clk_get_error;
> + }
> +
> + ret = msm_dss_clk_set_rate(mp->clk_config, mp->num_clk);
> + if (ret) {
> + DPU_ERROR("failed to set clock rate, ret=%d\n", ret);
> + goto clk_rate_error;
> + }
> +
> + dpu_mdss->base.dev = dev;
> + dpu_mdss->base.funcs = &mdss_funcs;
> +
> + ret = _dpu_mdss_irq_domain_add(dpu_mdss);
> + if (ret)
> + goto irq_domain_error;
> +
> + ret = devm_request_irq(dev->dev, platform_get_irq(pdev, 0),
> + dpu_mdss_irq, 0, "dpu_mdss_isr", dpu_mdss);
> + if (ret) {
> + DPU_ERROR("failed to init irq: %d\n", ret);
> + goto irq_error;
> + }
> +
> + pm_runtime_enable(dev->dev);
> +
> + pm_runtime_get_sync(dev->dev);
> + _dpu_mdss_hw_rev_init(dpu_mdss);
> + pm_runtime_put_sync(dev->dev);
> +
> + priv->mdss = &dpu_mdss->base;
> +
> + return ret;
> +
> +irq_error:
> + _dpu_mdss_irq_domain_fini(dpu_mdss);
> +irq_domain_error:
> +clk_rate_error:
> + msm_dss_put_clk(mp->clk_config, mp->num_clk);
> +clk_get_error:
> + devm_kfree(&pdev->dev, mp->clk_config);
> +clk_parse_err:
> + if (dpu_mdss->mmio)
> + msm_iounmap(pdev, dpu_mdss->mmio);
> + dpu_mdss->mmio = NULL;
> +error:
> + devm_kfree(dev->dev, dpu_mdss);
> + return ret;
> +}
> diff --git a/drivers/gpu/drm/msm/dpu_io_util.c b/drivers/gpu/drm/msm/dpu_io_util.c
> index a18bc99..efa06a8 100644
> --- a/drivers/gpu/drm/msm/dpu_io_util.c
> +++ b/drivers/gpu/drm/msm/dpu_io_util.c
> @@ -448,6 +448,61 @@ int msm_dss_enable_clk(struct dss_clk *clk_arry, int num_clk, int enable)
> } /* msm_dss_enable_clk */
> EXPORT_SYMBOL(msm_dss_enable_clk);
>
> +int msm_dss_parse_clock(struct platform_device *pdev,
> + struct dss_module_power *mp)
> +{
> + u32 i = 0, rc = 0;
> + const char *clock_name;
> + u32 clock_rate = 0;
> + u32 clock_max_rate = 0;
> + int num_clk = 0;
> +
> + if (!pdev || !mp) {
> + pr_err("invalid input param pdev:%pK mp:%pK\n", pdev, mp);
I don't think a log is necessary here. WARN if you must, but I would just
return.
> + return -EINVAL;
> + }
> +
> + mp->num_clk = 0;
> + num_clk = of_property_count_strings(pdev->dev.of_node, "clock-names");
> + if (num_clk <= 0) {
> + pr_err/*debug*/("clocks are not defined\n");
Ouch - please don't use internal comments */
> + goto clk_err;
rc is 0 this point - you can just return 0 here (or error if you choose).
> + }
> +
> + mp->num_clk = num_clk;
> + mp->clk_config = devm_kzalloc(&pdev->dev,
> + sizeof(struct dss_clk) * num_clk, GFP_KERNEL);
> + if (!mp->clk_config) {
> + rc = -ENOMEM;
> + mp->num_clk = 0;
> + goto clk_err;
> + }
> +
> + for (i = 0; i < num_clk; i++) {
> + of_property_read_string_index(pdev->dev.of_node, "clock-names",
> + i, &clock_name);
> + strlcpy(mp->clk_config[i].clk_name, clock_name,
> + sizeof(mp->clk_config[i].clk_name));
> +
> + of_property_read_u32_index(pdev->dev.of_node, "clock-rate",
> + i, &clock_rate);
> + mp->clk_config[i].rate = clock_rate;
> +
> + if (!clock_rate)
> + mp->clk_config[i].type = DSS_CLK_AHB;
> + else
> + mp->clk_config[i].type = DSS_CLK_PCLK;
> +
> + clock_max_rate = 0;
> + of_property_read_u32_index(pdev->dev.of_node, "clock-max-rate",
> + i, &clock_max_rate);
> + mp->clk_config[i].max_rate = clock_max_rate;
> + }
> +
> +clk_err:
> + return rc;
> +}
> +EXPORT_SYMBOL(msm_dss_parse_clock);
This needs to be exported? What other upstream modules would use this?
> int dpu_i2c_byte_read(struct i2c_client *client, uint8_t slave_addr,
> uint8_t reg_offset, uint8_t *read_buf)
> diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
> index 5d8f1b6..a0e73ea 100644
> --- a/drivers/gpu/drm/msm/msm_drv.c
> +++ b/drivers/gpu/drm/msm/msm_drv.c
> @@ -503,7 +503,18 @@ static int msm_drm_init(struct device *dev, struct drm_driver *drv)
> ddev->dev_private = priv;
> priv->dev = ddev;
>
> - ret = mdp5_mdss_init(ddev);
> + switch (get_mdp_ver(pdev)) {
> + case KMS_MDP5:
> + ret = mdp5_mdss_init(ddev);
> + break;
> + case KMS_DPU:
> + ret = dpu_mdss_init(ddev);
> + break;
> + default:
> + ret = 0;
> + break;
> + }
> +
> if (ret)
> goto mdss_init_fail;
>
> @@ -1539,12 +1550,13 @@ static int add_display_components(struct device *dev,
> int ret;
>
> /*
> - * MDP5 based devices don't have a flat hierarchy. There is a top level
> - * parent: MDSS, and children: MDP5, DSI, HDMI, eDP etc. Populate the
> - * children devices, find the MDP5 node, and then add the interfaces
> - * to our components list.
> + * MDP5/DPU based devices don't have a flat hierarchy. There is a top
> + * level parent: MDSS, and children: MDP5/DPU, DSI, HDMI, eDP etc.
> + * Populate the children devices, find the MDP5/DPU node, and then add
> + * the interfaces to our components list.
> */
> - if (of_device_is_compatible(dev->of_node, "qcom,mdss")) {
> + if (of_device_is_compatible(dev->of_node, "qcom,mdss") ||
> + of_device_is_compatible(dev->of_node, "qcom,dpu-mdss")) {
> ret = of_platform_populate(dev->of_node, NULL, NULL, dev);
> if (ret) {
> dev_err(dev, "failed to populate children devices\n");
> @@ -1686,7 +1698,7 @@ static int msm_pdev_remove(struct platform_device *pdev)
> { .compatible = "qcom,mdp4", .data = (void *)KMS_MDP4 },
> { .compatible = "qcom,mdss", .data = (void *)KMS_MDP5 },
> #ifdef CONFIG_DRM_MSM_DPU
> - { .compatible = "qcom,dpu-kms", .data = (void *)KMS_DPU },
> + { .compatible = "qcom,dpu-mdss", .data = (void *)KMS_DPU },
> #endif
> {}
> };
> diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h
> index 90a2521..e8e5e73 100644
> --- a/drivers/gpu/drm/msm/msm_drv.h
> +++ b/drivers/gpu/drm/msm/msm_drv.h
> @@ -381,7 +381,7 @@ struct msm_drm_private {
> /* subordinate devices, if present: */
> struct platform_device *gpu_pdev;
>
> - /* top level MDSS wrapper device (for MDP5 only) */
> + /* top level MDSS wrapper device (for MDP5/DPU only) */
> struct msm_mdss *mdss;
>
> /* possibly this should be in the kms component, but it is
> diff --git a/drivers/gpu/drm/msm/msm_kms.h b/drivers/gpu/drm/msm/msm_kms.h
> index 9a7bc7d..8f50613 100644
> --- a/drivers/gpu/drm/msm/msm_kms.h
> +++ b/drivers/gpu/drm/msm/msm_kms.h
> @@ -145,6 +145,8 @@ struct msm_mdss {
>
> int mdp5_mdss_init(struct drm_device *dev);
>
> +int dpu_mdss_init(struct drm_device *dev);
> +
> /**
> * Mode Set Utility Functions
> */
> diff --git a/include/linux/dpu_io_util.h b/include/linux/dpu_io_util.h
> index 7c73899..45e606f 100644
> --- a/include/linux/dpu_io_util.h
> +++ b/include/linux/dpu_io_util.h
> @@ -104,6 +104,8 @@ int msm_dss_config_vreg(struct device *dev, struct dss_vreg *in_vreg,
> void msm_dss_put_clk(struct dss_clk *clk_arry, int num_clk);
> int msm_dss_clk_set_rate(struct dss_clk *clk_arry, int num_clk);
> int msm_dss_enable_clk(struct dss_clk *clk_arry, int num_clk, int enable);
> +int msm_dss_parse_clock(struct platform_device *pdev,
> + struct dss_module_power *mp);
>
> int dpu_i2c_byte_read(struct i2c_client *client, uint8_t slave_addr,
> uint8_t reg_offset, uint8_t *read_buf);
--
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel
