| Message ID | 20170704175117.8636-1-robdclark@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
On Tue, Jul 4, 2017 at 7:51 PM, Rob Clark <robdclark@gmail.com> wrote: > The goal here is to support inheriting a display setup by bootloader, > although there may also be some non-display related use-cases. > > Rough idea is to add a flag for clks and power domains that might > already be enabled when kernel starts, and make corresponding fixups > to clk enable/prepare_count and power-domain state so that these are > not automatically disabled late in boot. > > If bootloader is enabling display, and kernel is using efifb before > real display driver is loaded (potentially from kernel module after > userspace starts, in a typical distro kernel), we don't want to kill > the clocks and power domains that are used by the display before > userspace starts. > > Second part will be (*waves hands*) for drm/msm to check if display > related clocks are enabled when it is loaded, and if so use drm > atomic framework's hooks to read back hw state to sync existing > display state w/ software state, and skip the initial clk_enable. > Therefore inheriting the enable done by bootloader. > > Obviously this should be split up into multiple patches and many > TODOs addressed. But I guess this is enough for now to start > discussing the approach, and in particular how drm and clock/pd > drivers work together to handle handover from bootloader. > > The CLK_INHERIT_BOOTLOADER and related gsdc flag should only be set > on leaf nodes. Cfr. Documentation/devicetree/bindings/display/simple-framebuffer.txt? Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds -- To unsubscribe from this list: send the line "unsubscribe linux-clk" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Jul 4, 2017 at 2:45 PM, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > On Tue, Jul 4, 2017 at 7:51 PM, Rob Clark <robdclark@gmail.com> wrote: >> The goal here is to support inheriting a display setup by bootloader, >> although there may also be some non-display related use-cases. >> >> Rough idea is to add a flag for clks and power domains that might >> already be enabled when kernel starts, and make corresponding fixups >> to clk enable/prepare_count and power-domain state so that these are >> not automatically disabled late in boot. >> >> If bootloader is enabling display, and kernel is using efifb before >> real display driver is loaded (potentially from kernel module after >> userspace starts, in a typical distro kernel), we don't want to kill >> the clocks and power domains that are used by the display before >> userspace starts. >> >> Second part will be (*waves hands*) for drm/msm to check if display >> related clocks are enabled when it is loaded, and if so use drm >> atomic framework's hooks to read back hw state to sync existing >> display state w/ software state, and skip the initial clk_enable. >> Therefore inheriting the enable done by bootloader. >> >> Obviously this should be split up into multiple patches and many >> TODOs addressed. But I guess this is enough for now to start >> discussing the approach, and in particular how drm and clock/pd >> drivers work together to handle handover from bootloader. >> >> The CLK_INHERIT_BOOTLOADER and related gsdc flag should only be set >> on leaf nodes. > > Cfr. Documentation/devicetree/bindings/display/simple-framebuffer.txt? > which afaict doesn't help for efifb.. which we have no bindings for (and would seem like a strange thing to have bindings for) BR, -R -- To unsubscribe from this list: send the line "unsubscribe linux-clk" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 07/04/2017 11:21 PM, Rob Clark wrote: > The goal here is to support inheriting a display setup by bootloader, > although there may also be some non-display related use-cases. > > Rough idea is to add a flag for clks and power domains that might > already be enabled when kernel starts, and make corresponding fixups > to clk enable/prepare_count and power-domain state so that these are > not automatically disabled late in boot. > > If bootloader is enabling display, and kernel is using efifb before > real display driver is loaded (potentially from kernel module after > userspace starts, in a typical distro kernel), we don't want to kill > the clocks and power domains that are used by the display before > userspace starts. > > Second part will be (*waves hands*) for drm/msm to check if display > related clocks are enabled when it is loaded, and if so use drm > atomic framework's hooks to read back hw state to sync existing > display state w/ software state, and skip the initial clk_enable. > Therefore inheriting the enable done by bootloader. > > Obviously this should be split up into multiple patches and many > TODOs addressed. But I guess this is enough for now to start > discussing the approach, and in particular how drm and clock/pd > drivers work together to handle handover from bootloader. > > The CLK_INHERIT_BOOTLOADER and related gsdc flag should only be set > on leaf nodes. > --- > A bit hacky right now, but display survives clk_disable_unused() > and genpd_power_off_unused(). It hangs just after that late in > boot, which I'm still debugging (might be unrelated shenanigans). > And haven't started on the drm/msm side of this. But I figured > it was half baked enough to send out for comments/ideas, or to > see if anyone had some different idea about how to solve this. Another RFC proposed around to handle similar situations is https://lkml.org/lkml/2017/6/28/188 That one though, I guess deals with only regulator supplies for now.
On Tue, Jul 4, 2017 at 11:27 PM, Rajendra Nayak <rnayak@codeaurora.org> wrote: > > > On 07/04/2017 11:21 PM, Rob Clark wrote: >> The goal here is to support inheriting a display setup by bootloader, >> although there may also be some non-display related use-cases. >> >> Rough idea is to add a flag for clks and power domains that might >> already be enabled when kernel starts, and make corresponding fixups >> to clk enable/prepare_count and power-domain state so that these are >> not automatically disabled late in boot. >> >> If bootloader is enabling display, and kernel is using efifb before >> real display driver is loaded (potentially from kernel module after >> userspace starts, in a typical distro kernel), we don't want to kill >> the clocks and power domains that are used by the display before >> userspace starts. >> >> Second part will be (*waves hands*) for drm/msm to check if display >> related clocks are enabled when it is loaded, and if so use drm >> atomic framework's hooks to read back hw state to sync existing >> display state w/ software state, and skip the initial clk_enable. >> Therefore inheriting the enable done by bootloader. >> >> Obviously this should be split up into multiple patches and many >> TODOs addressed. But I guess this is enough for now to start >> discussing the approach, and in particular how drm and clock/pd >> drivers work together to handle handover from bootloader. >> >> The CLK_INHERIT_BOOTLOADER and related gsdc flag should only be set >> on leaf nodes. >> --- >> A bit hacky right now, but display survives clk_disable_unused() >> and genpd_power_off_unused(). It hangs just after that late in >> boot, which I'm still debugging (might be unrelated shenanigans). >> And haven't started on the drm/msm side of this. But I figured >> it was half baked enough to send out for comments/ideas, or to >> see if anyone had some different idea about how to solve this. > > Another RFC proposed around to handle similar situations is > https://lkml.org/lkml/2017/6/28/188 > > That one though, I guess deals with only regulator supplies for now. Interesting.. although I do prefer it not being a static thing. Ie. depending on lk version and 'fastboot oem' params it may or may not light up the display. So detecting whether something is on at boot is, imo, important. (I haven't looked at regulators yet.. looks like it is all power domains, at least on 8x16, not sure about others..) BR, -R -- To unsubscribe from this list: send the line "unsubscribe linux-clk" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c index fc58c52..7c84e8d 100644 --- a/drivers/clk/clk.c +++ b/drivers/clk/clk.c @@ -796,6 +796,24 @@ static void clk_disable_unused_subtree(struct clk_core *core) clk_core_disable_unprepare(core->parent); } +/* clock and it's parents are already prepared/enabled from bootloader, + * so simply record the fact. + */ +static void __clk_inherit_enabled(struct clk_core *core) +{ + core->enable_count++; + core->prepare_count++; +if (core->hw->init->name) // XXX hack.. something funny with xo/xo_board.. + if (core->parent) + __clk_inherit_enabled(core->parent); +} + +void clk_inherit_enabled(struct clk *clk) +{ + __clk_inherit_enabled(clk->core); +} +EXPORT_SYMBOL_GPL(clk_inherit_enabled); + static bool clk_ignore_unused; static int __init clk_ignore_unused_setup(char *__unused) { diff --git a/drivers/clk/qcom/common.c b/drivers/clk/qcom/common.c index d523991..90b698c 100644 --- a/drivers/clk/qcom/common.c +++ b/drivers/clk/qcom/common.c @@ -11,6 +11,7 @@ * GNU General Public License for more details. */ +#include <linux/clk.h> #include <linux/export.h> #include <linux/module.h> #include <linux/regmap.h> @@ -258,6 +259,33 @@ int qcom_cc_really_probe(struct platform_device *pdev, if (ret) return ret; + /* Check which of clocks that we inherit state from bootloader + * are enabled, and fixup enable/prepare state (as well as that + * of it's parents). + * + * TODO can we assume that parents coming from another clk + * driver are already registered? + */ + for (i = 0; i < num_clks; i++) { + struct clk_hw *hw; + + if (!rclks[i]) + continue; + + hw = &rclks[i]->hw; + + if (!(hw->init->flags & CLK_INHERIT_BOOTLOADER)) + continue; + + if (!clk_is_enabled_regmap(hw)) + continue; + + dev_dbg(dev, "%s is enabled from bootloader!\n", + hw->init->name); + + clk_inherit_enabled(hw->clk); + } + reset = &cc->reset; reset->rcdev.of_node = dev->of_node; reset->rcdev.ops = &qcom_reset_ops; diff --git a/drivers/clk/qcom/gcc-msm8916.c b/drivers/clk/qcom/gcc-msm8916.c index 628e6ca..eca64f7 100644 --- a/drivers/clk/qcom/gcc-msm8916.c +++ b/drivers/clk/qcom/gcc-msm8916.c @@ -2467,7 +2467,7 @@ static struct clk_branch gcc_mdss_ahb_clk = { "pcnoc_bfdcd_clk_src", }, .num_parents = 1, - .flags = CLK_SET_RATE_PARENT, + .flags = CLK_SET_RATE_PARENT | CLK_INHERIT_BOOTLOADER, .ops = &clk_branch2_ops, }, }, @@ -2484,7 +2484,7 @@ static struct clk_branch gcc_mdss_axi_clk = { "system_noc_bfdcd_clk_src", }, .num_parents = 1, - .flags = CLK_SET_RATE_PARENT, + .flags = CLK_SET_RATE_PARENT | CLK_INHERIT_BOOTLOADER, .ops = &clk_branch2_ops, }, }, @@ -2501,7 +2501,7 @@ static struct clk_branch gcc_mdss_byte0_clk = { "byte0_clk_src", }, .num_parents = 1, - .flags = CLK_SET_RATE_PARENT, + .flags = CLK_SET_RATE_PARENT | CLK_INHERIT_BOOTLOADER, .ops = &clk_branch2_ops, }, }, @@ -2518,7 +2518,7 @@ static struct clk_branch gcc_mdss_esc0_clk = { "esc0_clk_src", }, .num_parents = 1, - .flags = CLK_SET_RATE_PARENT, + .flags = CLK_SET_RATE_PARENT | CLK_INHERIT_BOOTLOADER, .ops = &clk_branch2_ops, }, }, @@ -2535,7 +2535,7 @@ static struct clk_branch gcc_mdss_mdp_clk = { "mdp_clk_src", }, .num_parents = 1, - .flags = CLK_SET_RATE_PARENT, + .flags = CLK_SET_RATE_PARENT | CLK_INHERIT_BOOTLOADER, .ops = &clk_branch2_ops, }, }, @@ -2552,7 +2552,7 @@ static struct clk_branch gcc_mdss_pclk0_clk = { "pclk0_clk_src", }, .num_parents = 1, - .flags = CLK_SET_RATE_PARENT, + .flags = CLK_SET_RATE_PARENT | CLK_INHERIT_BOOTLOADER, .ops = &clk_branch2_ops, }, }, @@ -2569,7 +2569,7 @@ static struct clk_branch gcc_mdss_vsync_clk = { "vsync_clk_src", }, .num_parents = 1, - .flags = CLK_SET_RATE_PARENT, + .flags = CLK_SET_RATE_PARENT | CLK_INHERIT_BOOTLOADER, .ops = &clk_branch2_ops, }, }, @@ -3059,6 +3059,7 @@ static struct gdsc mdss_gdsc = { .name = "mdss", }, .pwrsts = PWRSTS_OFF_ON, + .flags = INHERIT_BL, }; static struct gdsc jpeg_gdsc = { diff --git a/drivers/clk/qcom/gdsc.c b/drivers/clk/qcom/gdsc.c index a4f3580..a1cf482 100644 --- a/drivers/clk/qcom/gdsc.c +++ b/drivers/clk/qcom/gdsc.c @@ -331,6 +331,15 @@ int gdsc_register(struct gdsc_desc *desc, if (ret) return ret; data->domains[i] = &scs[i]->pd; + + // TODO how do we tell if it is actually active? If it is not + // really active and we do this, it won't be enabled when + // driver wants it! + // + // TODO figure out how powerdomain state is refcnted so we + // can properly tell genpd that the domain is in use. + if (scs[i]->flags & INHERIT_BL) + data->domains[i]->flags |= GENPD_FLAG_ALWAYS_ON; } /* Add subdomains */ diff --git a/drivers/clk/qcom/gdsc.h b/drivers/clk/qcom/gdsc.h index 3964834..3b5e64b 100644 --- a/drivers/clk/qcom/gdsc.h +++ b/drivers/clk/qcom/gdsc.h @@ -53,6 +53,7 @@ struct gdsc { #define VOTABLE BIT(0) #define CLAMP_IO BIT(1) #define HW_CTRL BIT(2) +#define INHERIT_BL BIT(3) struct reset_controller_dev *rcdev; unsigned int *resets; unsigned int reset_count; diff --git a/include/linux/clk-provider.h b/include/linux/clk-provider.h index a428aec..7790382 100644 --- a/include/linux/clk-provider.h +++ b/include/linux/clk-provider.h @@ -35,6 +35,7 @@ #define CLK_IS_CRITICAL BIT(11) /* do not gate, ever */ /* parents need enable during gate/ungate, set rate and re-parent */ #define CLK_OPS_PARENT_ENABLE BIT(12) +#define CLK_INHERIT_BOOTLOADER BIT(13) /* clk may be enabled from bootloader */ struct clk; struct clk_hw; diff --git a/include/linux/clk.h b/include/linux/clk.h index 024cd07..43003e9 100644 --- a/include/linux/clk.h +++ b/include/linux/clk.h @@ -295,6 +295,15 @@ int clk_enable(struct clk *clk); void clk_disable(struct clk *clk); /** + * clk_inherit_enabled - update the enable/prepare count of a clock and it's + * parents for clock enabled by bootloader. + * + * Intended to be used by clock drivers to inform the clk core of a clock + * that is already running. + */ +void clk_inherit_enabled(struct clk *clk); + +/** * clk_get_rate - obtain the current clock rate (in Hz) for a clock source. * This is only valid once the clock source has been enabled. * @clk: clock source