| Message ID | 20180626044543.14992-1-spanda@codeaurora.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi, On Mon, Jun 25, 2018 at 9:45 PM, Sandeep Panda <spanda@codeaurora.org> wrote: > From: Abhinav Kumar <abhinavk@codeaurora.org> > > Setting the DSI PLL src in probe doesn't provide the clock > driver sufficient time to reclaim unused clock resources > from coreboot resulting in warnings from clock driver. > > Move the DSI PLL src setting to modeset_init() so that the > clock driver can claim unused display clock resources before > the display driver requests for them again. IMHO this is a bad design. Sean and Stephen can feel free to override me, but I think the clock driver should be improved to handle this case and not require the clock to get disabled before Linux enables it. -Doug
On Tue, Jun 26, 2018 at 11:55 AM, Doug Anderson <dianders@chromium.org> wrote: > Hi, > > On Mon, Jun 25, 2018 at 9:45 PM, Sandeep Panda <spanda@codeaurora.org> wrote: >> From: Abhinav Kumar <abhinavk@codeaurora.org> >> >> Setting the DSI PLL src in probe doesn't provide the clock >> driver sufficient time to reclaim unused clock resources >> from coreboot resulting in warnings from clock driver. >> >> Move the DSI PLL src setting to modeset_init() so that the >> clock driver can claim unused display clock resources before >> the display driver requests for them again. > > IMHO this is a bad design. Sean and Stephen can feel free to override > me, but I think the clock driver should be improved to handle this > case and not require the clock to get disabled before Linux enables > it. > I experimented with it a while back[1] (in this case w/ lk lighting up the display). In that case I needed both the clk and gdsc code to realize that clks/gdsc's were on at boot, and fixup the refcnt of the clks (and parent clocks and so on). And then when probed the display driver would check if clocks were enabled to decide to readback the state from the hw. (Maybe you can short-circuit some of that if you only care about DSI panels with a single fixed resolution, but as soon as external displays come into the picture you can't assume so much about the hw state.) BR, -R [1] https://github.com/freedreno/kernel-msm/commits/display-handover
Hi, On Tue, Jun 26, 2018 at 10:27 AM, Rob Clark <robdclark@gmail.com> wrote: > On Tue, Jun 26, 2018 at 11:55 AM, Doug Anderson <dianders@chromium.org> wrote: >> Hi, >> >> On Mon, Jun 25, 2018 at 9:45 PM, Sandeep Panda <spanda@codeaurora.org> wrote: >>> From: Abhinav Kumar <abhinavk@codeaurora.org> >>> >>> Setting the DSI PLL src in probe doesn't provide the clock >>> driver sufficient time to reclaim unused clock resources >>> from coreboot resulting in warnings from clock driver. >>> >>> Move the DSI PLL src setting to modeset_init() so that the >>> clock driver can claim unused display clock resources before >>> the display driver requests for them again. >> >> IMHO this is a bad design. Sean and Stephen can feel free to override >> me, but I think the clock driver should be improved to handle this >> case and not require the clock to get disabled before Linux enables >> it. >> > > I experimented with it a while back[1] (in this case w/ lk lighting up > the display). In that case I needed both the clk and gdsc code to > realize that clks/gdsc's were on at boot, and fixup the refcnt of the > clks (and parent clocks and so on). And then when probed the display > driver would check if clocks were enabled to decide to readback the > state from the hw. (Maybe you can short-circuit some of that if you > only care about DSI panels with a single fixed resolution, but as soon > as external displays come into the picture you can't assume so much > about the hw state.) > > BR, > -R > > [1] https://github.com/freedreno/kernel-msm/commits/display-handover It seems like something like that is the right real solution here, but I guess someone needs to step up and work on it. ...in general I'm wary of any patch that will not work when "clk_ignore_unused" is passed as a command line parameter. It's useful to be able to use this command line parameter for debugging sometimes and it would be unfortunate if doing so spewed a bunch of extra warnings. On a side node, it appears that even without ${SUBJECT} patch the warnings seem to have gone away with the latest stack of patches I've been testing. The warnings don't even come back with "clk_ignore_unused". I haven't personally dug into why, but I figured I'd mention it. -Doug
On Wed, Jun 27, 2018 at 2:48 PM, Doug Anderson <dianders@chromium.org> wrote: > Hi, > > On Tue, Jun 26, 2018 at 10:27 AM, Rob Clark <robdclark@gmail.com> wrote: >> On Tue, Jun 26, 2018 at 11:55 AM, Doug Anderson <dianders@chromium.org> wrote: >>> Hi, >>> >>> On Mon, Jun 25, 2018 at 9:45 PM, Sandeep Panda <spanda@codeaurora.org> wrote: >>>> From: Abhinav Kumar <abhinavk@codeaurora.org> >>>> >>>> Setting the DSI PLL src in probe doesn't provide the clock >>>> driver sufficient time to reclaim unused clock resources >>>> from coreboot resulting in warnings from clock driver. >>>> >>>> Move the DSI PLL src setting to modeset_init() so that the >>>> clock driver can claim unused display clock resources before >>>> the display driver requests for them again. >>> >>> IMHO this is a bad design. Sean and Stephen can feel free to override >>> me, but I think the clock driver should be improved to handle this >>> case and not require the clock to get disabled before Linux enables >>> it. >>> >> >> I experimented with it a while back[1] (in this case w/ lk lighting up >> the display). In that case I needed both the clk and gdsc code to >> realize that clks/gdsc's were on at boot, and fixup the refcnt of the >> clks (and parent clocks and so on). And then when probed the display >> driver would check if clocks were enabled to decide to readback the >> state from the hw. (Maybe you can short-circuit some of that if you >> only care about DSI panels with a single fixed resolution, but as soon >> as external displays come into the picture you can't assume so much >> about the hw state.) >> >> BR, >> -R >> >> [1] https://github.com/freedreno/kernel-msm/commits/display-handover > > It seems like something like that is the right real solution here, but > I guess someone needs to step up and work on it. > yeah, sadly I haven't found time to revisit it (kinda been short on time to work on display/kernel side of things.. it would be helpful if qcom would stop coming out with new gpu's so often :-P) But I am pretty sure if we want a general solution that can also deal w/ splash screen on hot-pluggable display, and not just the simpler case of fixed resolution dsi panels, it is going to require cooperation between clk/gdsc and display driver. > ...in general I'm wary of any patch that will not work when > "clk_ignore_unused" is passed as a command line parameter. It's > useful to be able to use this command line parameter for debugging > sometimes and it would be unfortunate if doing so spewed a bunch of > extra warnings. for "full distro" config where all the display drivers are built as modules and prior to real display driver loading you have kernel inheriting the display via efifb or simplefb, I think we need to be able to flag certain clocks and power domains (and on other platforms, perhaps regulators) as "if the bootloader left this on, keep it that way".. I guess for simple-fb we could perhaps solve it by attaching power domains and clks to the simple-fb node, but that doesn't really work for efifb.. BR, -R > > > On a side node, it appears that even without ${SUBJECT} patch the > warnings seem to have gone away with the latest stack of patches I've > been testing. The warnings don't even come back with > "clk_ignore_unused". I haven't personally dug into why, but I figured > I'd mention it. > > > -Doug
A brief update on this topic: The DSI clock warnings are addressed after adding runtime_pm support to DPU driver [1]. MDSS GDSC is used as genpd w/ above series and is requested by parent MDSS device on behalf of all child devices (like DPU, DSI etc). Before adding the runtime_pm support, DSI driver probing was happening before MDSS GDSC initialization and is why the RCG clock configuration for byte and pixel clocks was not consumed by hardware leading to the warnings. Regarding display handover from bootloader to Linux display drivers (referred as continuous splash feature), is supported downstream for DSI displays (where splash screen in handed over to kernel drivers w/o re-init'ing the hardware) and needs coordination (as pointed out by Rob) b/t display, clock, gdsc and iommu teams. But continuous splash for hot-pluggable displays is not attempted so far. We will discuss among the teams to see if continuous slash changes (present downstream) in conjunction w/ Rob's work can make their way upstream. [1] https://lists.freedesktop.org/archives/freedreno/2018-May/002502.html Thanks, Rajesh On 2018-06-28 06:05, Rob Clark wrote: > On Wed, Jun 27, 2018 at 2:48 PM, Doug Anderson <dianders@chromium.org> > wrote: >> Hi, >> >> On Tue, Jun 26, 2018 at 10:27 AM, Rob Clark <robdclark@gmail.com> >> wrote: >>> On Tue, Jun 26, 2018 at 11:55 AM, Doug Anderson >>> <dianders@chromium.org> wrote: >>>> Hi, >>>> >>>> On Mon, Jun 25, 2018 at 9:45 PM, Sandeep Panda >>>> <spanda@codeaurora.org> wrote: >>>>> From: Abhinav Kumar <abhinavk@codeaurora.org> >>>>> >>>>> Setting the DSI PLL src in probe doesn't provide the clock >>>>> driver sufficient time to reclaim unused clock resources >>>>> from coreboot resulting in warnings from clock driver. >>>>> >>>>> Move the DSI PLL src setting to modeset_init() so that the >>>>> clock driver can claim unused display clock resources before >>>>> the display driver requests for them again. >>>> >>>> IMHO this is a bad design. Sean and Stephen can feel free to >>>> override >>>> me, but I think the clock driver should be improved to handle this >>>> case and not require the clock to get disabled before Linux enables >>>> it. >>>> >>> >>> I experimented with it a while back[1] (in this case w/ lk lighting >>> up >>> the display). In that case I needed both the clk and gdsc code to >>> realize that clks/gdsc's were on at boot, and fixup the refcnt of the >>> clks (and parent clocks and so on). And then when probed the display >>> driver would check if clocks were enabled to decide to readback the >>> state from the hw. (Maybe you can short-circuit some of that if you >>> only care about DSI panels with a single fixed resolution, but as >>> soon >>> as external displays come into the picture you can't assume so much >>> about the hw state.) >>> >>> BR, >>> -R >>> >>> [1] https://github.com/freedreno/kernel-msm/commits/display-handover >> >> It seems like something like that is the right real solution here, but >> I guess someone needs to step up and work on it. >> > > yeah, sadly I haven't found time to revisit it (kinda been short on > time to work on display/kernel side of things.. it would be helpful if > qcom would stop coming out with new gpu's so often :-P) > > But I am pretty sure if we want a general solution that can also deal > w/ splash screen on hot-pluggable display, and not just the simpler > case of fixed resolution dsi panels, it is going to require > cooperation between clk/gdsc and display driver. > >> ...in general I'm wary of any patch that will not work when >> "clk_ignore_unused" is passed as a command line parameter. It's >> useful to be able to use this command line parameter for debugging >> sometimes and it would be unfortunate if doing so spewed a bunch of >> extra warnings. > > for "full distro" config where all the display drivers are built as > modules and prior to real display driver loading you have kernel > inheriting the display via efifb or simplefb, I think we need to be > able to flag certain clocks and power domains (and on other platforms, > perhaps regulators) as "if the bootloader left this on, keep it that > way".. I guess for simple-fb we could perhaps solve it by attaching > power domains and clks to the simple-fb node, but that doesn't really > work for efifb.. > > BR, > -R > > >> >> >> On a side node, it appears that even without ${SUBJECT} patch the >> warnings seem to have gone away with the latest stack of patches I've >> been testing. The warnings don't even come back with >> "clk_ignore_unused". I haven't personally dug into why, but I figured >> I'd mention it. >> >> >> -Doug > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/dri-devel
diff --git a/drivers/gpu/drm/msm/dsi/dsi.c b/drivers/gpu/drm/msm/dsi/dsi.c index 9fb612c17a39..ea100f15ce0d 100644 --- a/drivers/gpu/drm/msm/dsi/dsi.c +++ b/drivers/gpu/drm/msm/dsi/dsi.c @@ -211,6 +211,12 @@ int msm_dsi_modeset_init(struct msm_dsi *msm_dsi, struct drm_device *dev, priv = dev->dev_private; msm_dsi->dev = dev; + ret = msm_dsi_manager_pll_setup(msm_dsi); + if (ret) { + dev_err(dev->dev, "failed to setup pll: %d\n", ret); + goto fail; + } + ret = msm_dsi_host_modeset_init(msm_dsi->host, dev); if (ret) { dev_err(dev->dev, "failed to modeset init host: %d\n", ret); diff --git a/drivers/gpu/drm/msm/dsi/dsi.h b/drivers/gpu/drm/msm/dsi/dsi.h index 356953010256..a9fed9ab17b2 100644 --- a/drivers/gpu/drm/msm/dsi/dsi.h +++ b/drivers/gpu/drm/msm/dsi/dsi.h @@ -101,6 +101,7 @@ void msm_dsi_manager_attach_dsi_device(int id, u32 device_flags); int msm_dsi_manager_register(struct msm_dsi *msm_dsi); void msm_dsi_manager_unregister(struct msm_dsi *msm_dsi); bool msm_dsi_manager_validate_current_config(u8 id); +int msm_dsi_manager_pll_setup(struct msm_dsi *msm_dsi); /* msm dsi */ static inline bool msm_dsi_device_connected(struct msm_dsi *msm_dsi) diff --git a/drivers/gpu/drm/msm/dsi/dsi_manager.c b/drivers/gpu/drm/msm/dsi/dsi_manager.c index 980b1bba8477..40da1229565e 100644 --- a/drivers/gpu/drm/msm/dsi/dsi_manager.c +++ b/drivers/gpu/drm/msm/dsi/dsi_manager.c @@ -72,25 +72,60 @@ static int dsi_mgr_parse_dual_dsi(struct device_node *np, int id) return 0; } -static int dsi_mgr_setup_components(int id) +static int msm_dsi_pll_configure(struct msm_dsi *msm_dsi, bool is_dual) { - struct msm_dsi *msm_dsi = dsi_mgr_get_dsi(id); - struct msm_dsi *other_dsi = dsi_mgr_get_other_dsi(id); + struct msm_dsi *other_dsi = dsi_mgr_get_other_dsi(msm_dsi->id); struct msm_dsi *clk_master_dsi = dsi_mgr_get_dsi(DSI_CLOCK_MASTER); struct msm_dsi *clk_slave_dsi = dsi_mgr_get_dsi(DSI_CLOCK_SLAVE); struct msm_dsi_pll *src_pll; - int ret; - - if (!IS_DUAL_DSI()) { - ret = msm_dsi_host_register(msm_dsi->host, true); - if (ret) - return ret; + int ret = 0; + if (!is_dual) { msm_dsi_phy_set_usecase(msm_dsi->phy, MSM_DSI_PHY_STANDALONE); src_pll = msm_dsi_phy_get_pll(msm_dsi->phy); if (IS_ERR(src_pll)) return PTR_ERR(src_pll); ret = msm_dsi_host_set_src_pll(msm_dsi->host, src_pll); + } else { + /* PLL0 is to drive both 2 DSI link clocks in Dual DSI mode. */ + msm_dsi_phy_set_usecase(clk_master_dsi->phy, + MSM_DSI_PHY_MASTER); + msm_dsi_phy_set_usecase(clk_slave_dsi->phy, + MSM_DSI_PHY_SLAVE); + src_pll = msm_dsi_phy_get_pll(clk_master_dsi->phy); + if (IS_ERR(src_pll)) + return PTR_ERR(src_pll); + ret = msm_dsi_host_set_src_pll(msm_dsi->host, src_pll); + if (ret) + return ret; + ret = msm_dsi_host_set_src_pll(other_dsi->host, src_pll); + } + + return ret; +} + +int msm_dsi_manager_pll_setup(struct msm_dsi *msm_dsi) +{ + int ret; + + if (!IS_DUAL_DSI()) + ret = msm_dsi_pll_configure(msm_dsi, true); + else + ret = msm_dsi_pll_configure(msm_dsi, false); + + return ret; +} + +static int dsi_mgr_setup_components(int id) +{ + struct msm_dsi *msm_dsi = dsi_mgr_get_dsi(id); + struct msm_dsi *other_dsi = dsi_mgr_get_other_dsi(id); + int ret; + + if (!IS_DUAL_DSI()) { + ret = msm_dsi_host_register(msm_dsi->host, true); + if (ret) + return ret; } else if (!other_dsi) { ret = 0; } else { @@ -111,19 +146,6 @@ static int dsi_mgr_setup_components(int id) ret = msm_dsi_host_register(master_link_dsi->host, true); if (ret) return ret; - - /* PLL0 is to drive both 2 DSI link clocks in Dual DSI mode. */ - msm_dsi_phy_set_usecase(clk_master_dsi->phy, - MSM_DSI_PHY_MASTER); - msm_dsi_phy_set_usecase(clk_slave_dsi->phy, - MSM_DSI_PHY_SLAVE); - src_pll = msm_dsi_phy_get_pll(clk_master_dsi->phy); - if (IS_ERR(src_pll)) - return PTR_ERR(src_pll); - ret = msm_dsi_host_set_src_pll(msm_dsi->host, src_pll); - if (ret) - return ret; - ret = msm_dsi_host_set_src_pll(other_dsi->host, src_pll); } return ret;