Message ID | 1414072090-11293-2-git-send-email-k.kozlowski@samsung.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Hi Krzysztof, On 10/23/2014 04:48 PM, Krzysztof Kozlowski wrote: > When resuming the system the power domain has to be powered on early so > any runtime PM aware devices could resume. > > This fixes following scenario reproduced on Exynos DRM: > 1. Power domain is off before suspending the system. > 2. System is suspended to RAM. > 3. Resuming starts. The Exynos DRM driver resume callback is called. > 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns > the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. > 5. The Exynos DSI driver calls pm_runtime_get. The driver runtime > resumes and this should turn LCD power domain on. > 6. Unfortunately the domain cannot be turned on because system resume is > in progress and genpd->prepared_count is positive. Just interesting, what value will be returned by pm_runtime_enabled() from any of your .resume() callback (for any device which belongs to some Generic PM domain)? I'm asking, because as I can see Runtime PM can be disabled from pm_genpd_prepare(). Thank you. Oh. I've just found that you might get this issue if you will try to do suspend when PM domain is ON ;) Any way, In my opinion, It might be better to fix pm_genpd_prepare() so it will not increment prepared_count when initial state of the GPD is GPD_STATE_POWER_OFF. Seems it's needed only in opposite case - when state of GPD has to be restored from pm_genpd_resume_noirq(). > > Steps to reproduce: > 1. Add runtime PM to Exynos DSI driver. > 2. Build Exynos DRM/FB without FRAMEBUFFER_CONSOLE. > 3. Enable the connector and screen (e.g. with modeset-vsync application). > 4. echo 3 > /sys/devices/platform/exynos-drm/graphics/fb0/blank > 5. echo mem > /sys/power/state > 6. Resume. > [ 77.712469] PM: early resume of devices complete after 3.854 msecs > [ 77.712739] exynos-dsi 11c80000.dsi: pm_genpd_resume() > [ 77.712758] exynos4-fimc 11800000.fimc: pm_genpd_resume() > [ 77.712774] exynos4-fimc 11810000.fimc: pm_genpd_resume() > [ 77.712787] exynos-drm-fimc 11820000.fimc: pm_genpd_resume() > [ 77.712802] exynos-drm-fimc 11830000.fimc: pm_genpd_resume() > [ 77.712815] s5p-mipi-csis 11880000.csis: pm_genpd_resume() > [ 77.712829] s5p-mipi-csis 11890000.csis: pm_genpd_resume() > [ 77.712843] exynos-fimc-lite 12390000.fimc-lite: pm_genpd_resume() > [ 77.712856] exynos-fimc-lite 123a0000.fimc-lite: pm_genpd_resume() > [ 77.713788] exynos4-fb 11c00000.fimd: pm_genpd_resume() > [ 77.713912] wake disabled for irq 184 > [ 77.713923] wake disabled for irq 185 > [ 77.714082] wake disabled for irq 173 > [ 77.715676] wake disabled for irq 176 > [ 77.718540] exynos4-fb 11c00000.fimd: pm_genpd_runtime_resume() > [ 77.718567] exynos4-fb 11c00000.fimd: state restore latency exceeded, new value 1708 ns > [ 77.718636] exynos-dsi 11c80000.dsi: pm_genpd_runtime_resume() > [ 77.892366] exynos-dsi 11c80000.dsi: PLL failed to stabilize > [ 77.892377] exynos-dsi 11c80000.dsi: failed to configure DSI PLL > [ 78.192168] exynos-dsi 11c80000.dsi: timeout waiting for reset > [ 78.211578] exynos-dsi 11c80000.dsi: waiting for bus lanes timed out > [ 78.307173] exynos-dsi 11c80000.dsi: xfer timed out: d1 00 (null) > [ 78.307190] panel_s6e8aa0 11c80000.dsi.0: error -110 reading dcs seq(0xd1) > [ 78.307199] panel_s6e8aa0 11c80000.dsi.0: read id failed > > Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> > --- > drivers/base/power/domain.c | 3 +-- > 1 file changed, 1 insertion(+), 2 deletions(-) > > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c > index 40bc2f4072cc..4fdfe404a04c 100644 > --- a/drivers/base/power/domain.c > +++ b/drivers/base/power/domain.c > @@ -179,8 +179,7 @@ static int __pm_genpd_poweron(struct generic_pm_domain *genpd) > } > finish_wait(&genpd->status_wait_queue, &wait); > > - if (genpd->status == GPD_STATE_ACTIVE > - || (genpd->prepared_count > 0 && genpd->suspend_power_off)) > + if (genpd->status == GPD_STATE_ACTIVE) > return 0; > > if (genpd->status != GPD_STATE_POWER_OFF) { > regards, -grygorii
On czw, 2014-10-23 at 19:20 +0300, Grygorii Strashko wrote: > Hi Krzysztof, > > On 10/23/2014 04:48 PM, Krzysztof Kozlowski wrote: > > When resuming the system the power domain has to be powered on early so > > any runtime PM aware devices could resume. > > > > This fixes following scenario reproduced on Exynos DRM: > > 1. Power domain is off before suspending the system. > > 2. System is suspended to RAM. > > 3. Resuming starts. The Exynos DRM driver resume callback is called. > > 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns > > the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. > > 5. The Exynos DSI driver calls pm_runtime_get. The driver runtime > > resumes and this should turn LCD power domain on. > > 6. Unfortunately the domain cannot be turned on because system resume is > > in progress and genpd->prepared_count is positive. > > Just interesting, what value will be returned by pm_runtime_enabled() > from any of your .resume() callback (for any device which belongs to > some Generic PM domain)? exynos_drm_resume: false exynos_dsi_enable: true Full backtrace leading to exynos_dsi_enable: [ 37.944830] ------------[ cut here ]------------ [ 37.944860] WARNING: CPU: 0 PID: 3125 at drivers/gpu/drm/exynos/exynos_drm_dsi.c:1360 exynos_dsi_dpms+0xc0/0x398() [ 37.944869] Modules linked in: [ 37.944883] CPU: 0 PID: 3125 Comm: bash Tainted: G W 3.17.0-next-20141020-00050-g844ed80678d5-dirty #479 [ 37.944923] [<c0013d64>] (unwind_backtrace) from [<c0010eb0>] (show_stack+0x10/0x14) [ 37.944949] [<c0010eb0>] (show_stack) from [<c04a37b0>] (dump_stack+0x70/0xbc) [ 37.944977] [<c04a37b0>] (dump_stack) from [<c0021420>] (warn_slowpath_common+0x64/0x88) [ 37.944992] [<c0021420>] (warn_slowpath_common) from [<c0021460>] (warn_slowpath_null+0x1c/0x24) [ 37.945011] [<c0021460>] (warn_slowpath_null) from [<c0268940>] (exynos_dsi_dpms+0xc0/0x398) [ 37.945024] [<c0268940>] (exynos_dsi_dpms) from [<c0263520>] (exynos_drm_encoder_commit+0x24/0x40) [ 37.945044] [<c0263520>] (exynos_drm_encoder_commit) from [<c023f0b4>] (drm_crtc_helper_set_mode+0x400/0x4e8) [ 37.945057] [<c023f0b4>] (drm_crtc_helper_set_mode) from [<c023f204>] (drm_helper_resume_force_mode+0x68/0x124) [ 37.945083] [<c023f204>] (drm_helper_resume_force_mode) from [<c0262b9c>] (exynos_drm_resume+0x80/0x90) [ 37.945100] [<c0262b9c>] (exynos_drm_resume) from [<c02832f8>] (platform_pm_resume+0x2c/0x4c) [ 37.945118] [<c02832f8>] (platform_pm_resume) from [<c028a320>] (dpm_run_callback.isra.7+0x2c/0x64) [ 37.945130] [<c028a320>] (dpm_run_callback.isra.7) from [<c028a404>] (device_resume+0xac/0x180) [ 37.945141] [<c028a404>] (device_resume) from [<c028b65c>] (dpm_resume+0xe8/0x20c) [ 37.945152] [<c028b65c>] (dpm_resume) from [<c028b91c>] (dpm_resume_end+0xc/0x18) [ 37.945175] [<c028b91c>] (dpm_resume_end) from [<c0054b48>] (suspend_devices_and_enter+0x230/0x3c8) [ 37.945190] [<c0054b48>] (suspend_devices_and_enter) from [<c0054f48>] (pm_suspend+0x268/0x29c) [ 37.945202] [<c0054f48>] (pm_suspend) from [<c0053a74>] (state_store+0x6c/0xbc) [ 37.945220] [<c0053a74>] (state_store) from [<c01d0890>] (kobj_attr_store+0x14/0x20) [ 37.945235] [<c01d0890>] (kobj_attr_store) from [<c011e56c>] (sysfs_kf_write+0x44/0x48) [ 37.945245] [<c011e56c>] (sysfs_kf_write) from [<c011dc10>] (kernfs_fop_write+0xc0/0x17c) [ 37.945268] [<c011dc10>] (kernfs_fop_write) from [<c00c7bfc>] (vfs_write+0xa0/0x1a8) [ 37.945282] [<c00c7bfc>] (vfs_write) from [<c00c7f14>] (SyS_write+0x40/0x8c) [ 37.945299] [<c00c7f14>] (SyS_write) from [<c000e6e0>] (ret_fast_syscall+0x0/0x30) [ 37.945307] ---[ end trace e930e0edfd9a5ad2 ]--- > I'm asking, because as I can see Runtime PM can be disabled from pm_genpd_prepare(). > > Thank you. > > Oh. I've just found that you might get this issue if you will try to do > suspend when PM domain is ON ;) > > Any way, In my opinion, It might be better to fix pm_genpd_prepare() so > it will not increment prepared_count when initial state of the GPD is > GPD_STATE_POWER_OFF. Seems it's needed only in opposite case - > when state of GPD has to be restored from pm_genpd_resume_noirq(). That sounds good but what about cases when the device will runtime resume during suspend (early at suspend)? Actually I seen this with framebuffer console. Right after starting suspend the console is flushed and poked which leads to powering on LCD. Best regards, Krzysztof
Krzysztof Kozlowski <k.kozlowski@samsung.com> writes: > When resuming the system the power domain has to be powered on early so > any runtime PM aware devices could resume. > > This fixes following scenario reproduced on Exynos DRM: > 1. Power domain is off before suspending the system. > 2. System is suspended to RAM. > 3. Resuming starts. The Exynos DRM driver resume callback is called. > 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns > the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. Dumb Q: if the device (and power domain) were off before (and during) suspend, why are they being resumed? Shouldn't the resume path restore things to the same state they were before suspend? Kevin
On ?ro, 2014-10-29 at 10:46 -0700, Kevin Hilman wrote: > Krzysztof Kozlowski <k.kozlowski@samsung.com> writes: > > > When resuming the system the power domain has to be powered on early so > > any runtime PM aware devices could resume. > > > > This fixes following scenario reproduced on Exynos DRM: > > 1. Power domain is off before suspending the system. > > 2. System is suspended to RAM. > > 3. Resuming starts. The Exynos DRM driver resume callback is called. > > 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns > > the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. > > Dumb Q: if the device (and power domain) were off before (and during) > suspend, why are they being resumed? > > Shouldn't the resume path restore things to the same state they were > before suspend? One could expect that... but the Exynos DRM driver behaves differently (and some other drivers also). In resume method it calls drm_helper_resume_force_mode() which forces restoring mode setting configuration. Apparently setting a mode needs DPMS on: static void exynos_drm_crtc_commit(struct drm_crtc *crtc) { ... exynos_drm_crtc_dpms(crtc, DRM_MODE_DPMS_ON); ... The previous DPMS status (status during suspend) is completely ignored here. I dunno why... maybe someone from DRM could share some thoughts? Best regards, Krzysztof
On 10/30/2014 08:36 AM, Krzysztof Kozlowski wrote: > On ?ro, 2014-10-29 at 10:46 -0700, Kevin Hilman wrote: >> Krzysztof Kozlowski <k.kozlowski@samsung.com> writes: >> >>> When resuming the system the power domain has to be powered on early so >>> any runtime PM aware devices could resume. >>> >>> This fixes following scenario reproduced on Exynos DRM: >>> 1. Power domain is off before suspending the system. >>> 2. System is suspended to RAM. >>> 3. Resuming starts. The Exynos DRM driver resume callback is called. >>> 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns >>> the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. >> Dumb Q: if the device (and power domain) were off before (and during) >> suspend, why are they being resumed? >> >> Shouldn't the resume path restore things to the same state they were >> before suspend? > One could expect that... but the Exynos DRM driver behaves differently > (and some other drivers also). In resume method it calls > drm_helper_resume_force_mode() which forces restoring mode setting > configuration. Apparently setting a mode needs DPMS on: > static void exynos_drm_crtc_commit(struct drm_crtc *crtc) > { > ... > exynos_drm_crtc_dpms(crtc, DRM_MODE_DPMS_ON); > ... > > The previous DPMS status (status during suspend) is completely ignored > here. Suspend callback switches off all connectors (thus all other devs in their pipeline) by calling dpms_off, in restore callback all devs are restored to their previous state by calling appropriate dpms. So I guess drm_helper_resume_force_mode() call at the end of resume is incorrect. On the other side it is present in many other drivers, so I am also little bit confused. Regards Andrzej
Andrzej Hajda <a.hajda@samsung.com> writes: > On 10/30/2014 08:36 AM, Krzysztof Kozlowski wrote: >> On ?ro, 2014-10-29 at 10:46 -0700, Kevin Hilman wrote: >>> Krzysztof Kozlowski <k.kozlowski@samsung.com> writes: >>> >>>> When resuming the system the power domain has to be powered on early so >>>> any runtime PM aware devices could resume. >>>> >>>> This fixes following scenario reproduced on Exynos DRM: >>>> 1. Power domain is off before suspending the system. >>>> 2. System is suspended to RAM. >>>> 3. Resuming starts. The Exynos DRM driver resume callback is called. >>>> 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns >>>> the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. >>> Dumb Q: if the device (and power domain) were off before (and during) >>> suspend, why are they being resumed? >>> >>> Shouldn't the resume path restore things to the same state they were >>> before suspend? >> One could expect that... but the Exynos DRM driver behaves differently >> (and some other drivers also). In resume method it calls >> drm_helper_resume_force_mode() which forces restoring mode setting >> configuration. Apparently setting a mode needs DPMS on: >> static void exynos_drm_crtc_commit(struct drm_crtc *crtc) >> { >> ... >> exynos_drm_crtc_dpms(crtc, DRM_MODE_DPMS_ON); >> ... >> >> The previous DPMS status (status during suspend) is completely ignored >> here. > > Suspend callback switches off all connectors (thus all other devs in > their pipeline) by calling dpms_off, > in restore callback all devs are restored to their previous state by > calling appropriate dpms. > So I guess drm_helper_resume_force_mode() call at the end of resume is > incorrect. Though I'm not terribly familiar with DRM, it seems incorrect because I expect resume to restore the state of things when suspend happened, not forcibly resume everything. > On the other side it is present in many other drivers, so I am also > little bit confused. Many other DRM drivers? or other drivers too? Kevin
On 11/03/2014 06:13 PM, Kevin Hilman wrote: > Andrzej Hajda <a.hajda@samsung.com> writes: > >> On 10/30/2014 08:36 AM, Krzysztof Kozlowski wrote: >>> On ?ro, 2014-10-29 at 10:46 -0700, Kevin Hilman wrote: >>>> Krzysztof Kozlowski <k.kozlowski@samsung.com> writes: >>>> >>>>> When resuming the system the power domain has to be powered on early so >>>>> any runtime PM aware devices could resume. >>>>> >>>>> This fixes following scenario reproduced on Exynos DRM: >>>>> 1. Power domain is off before suspending the system. >>>>> 2. System is suspended to RAM. >>>>> 3. Resuming starts. The Exynos DRM driver resume callback is called. >>>>> 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns >>>>> the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. >>>> Dumb Q: if the device (and power domain) were off before (and during) >>>> suspend, why are they being resumed? >>>> >>>> Shouldn't the resume path restore things to the same state they were >>>> before suspend? >>> One could expect that... but the Exynos DRM driver behaves differently >>> (and some other drivers also). In resume method it calls >>> drm_helper_resume_force_mode() which forces restoring mode setting >>> configuration. Apparently setting a mode needs DPMS on: >>> static void exynos_drm_crtc_commit(struct drm_crtc *crtc) >>> { >>> ... >>> exynos_drm_crtc_dpms(crtc, DRM_MODE_DPMS_ON); >>> ... >>> >>> The previous DPMS status (status during suspend) is completely ignored >>> here. >> >> Suspend callback switches off all connectors (thus all other devs in >> their pipeline) by calling dpms_off, >> in restore callback all devs are restored to their previous state by >> calling appropriate dpms. >> So I guess drm_helper_resume_force_mode() call at the end of resume is >> incorrect. > > Though I'm not terribly familiar with DRM, it seems incorrect because I > expect resume to restore the state of things when suspend happened, not > forcibly resume everything. > >> On the other side it is present in many other drivers, so I am also >> little bit confused. > > Many other DRM drivers? or other drivers too? If I understand GPD code right :) There is "small" problem :( Now if PM domain is OFF before suspend - it's prohibited to turn it on during suspending/resuming. commit 596ba34bcd2978ee9823cc1d84df230576f8ffb9 PM / Domains: System-wide transitions support for generic domains (v5) But, it is possible to have devices which supports few power states, like: on, sleep/low power, off (for example OMAP devices and also I saw this for some I2C devices - can recollect only that it was some sensor and touchscreen). In normal operational mode Runtime PM switches device between on and sleep/low power states, but during suspend device need to be woken up and reconfigured to off state. Also I found, that It looks like due to continuous refactoring the call of pm_generic_suspend_noirq(dev) was finally dropped from pm_genpd_suspend_noirq(), so .suspend_noirq() callback will never be called for devices from GPD. Seems, that is the problem which this patch tries to fix and looks like there are will be more such kind of report as GPD is become used widely. regards, -grygorii
diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c index 40bc2f4072cc..4fdfe404a04c 100644 --- a/drivers/base/power/domain.c +++ b/drivers/base/power/domain.c @@ -179,8 +179,7 @@ static int __pm_genpd_poweron(struct generic_pm_domain *genpd) } finish_wait(&genpd->status_wait_queue, &wait); - if (genpd->status == GPD_STATE_ACTIVE - || (genpd->prepared_count > 0 && genpd->suspend_power_off)) + if (genpd->status == GPD_STATE_ACTIVE) return 0; if (genpd->status != GPD_STATE_POWER_OFF) {
When resuming the system the power domain has to be powered on early so any runtime PM aware devices could resume. This fixes following scenario reproduced on Exynos DRM: 1. Power domain is off before suspending the system. 2. System is suspended to RAM. 3. Resuming starts. The Exynos DRM driver resume callback is called. 4. The Exynos DRM driver calls drm_helper_resume_force_mode which turns the screen on by calling exynos_dsi_dpms with DRM_MODE_DPMS_ON. 5. The Exynos DSI driver calls pm_runtime_get. The driver runtime resumes and this should turn LCD power domain on. 6. Unfortunately the domain cannot be turned on because system resume is in progress and genpd->prepared_count is positive. Steps to reproduce: 1. Add runtime PM to Exynos DSI driver. 2. Build Exynos DRM/FB without FRAMEBUFFER_CONSOLE. 3. Enable the connector and screen (e.g. with modeset-vsync application). 4. echo 3 > /sys/devices/platform/exynos-drm/graphics/fb0/blank 5. echo mem > /sys/power/state 6. Resume. [ 77.712469] PM: early resume of devices complete after 3.854 msecs [ 77.712739] exynos-dsi 11c80000.dsi: pm_genpd_resume() [ 77.712758] exynos4-fimc 11800000.fimc: pm_genpd_resume() [ 77.712774] exynos4-fimc 11810000.fimc: pm_genpd_resume() [ 77.712787] exynos-drm-fimc 11820000.fimc: pm_genpd_resume() [ 77.712802] exynos-drm-fimc 11830000.fimc: pm_genpd_resume() [ 77.712815] s5p-mipi-csis 11880000.csis: pm_genpd_resume() [ 77.712829] s5p-mipi-csis 11890000.csis: pm_genpd_resume() [ 77.712843] exynos-fimc-lite 12390000.fimc-lite: pm_genpd_resume() [ 77.712856] exynos-fimc-lite 123a0000.fimc-lite: pm_genpd_resume() [ 77.713788] exynos4-fb 11c00000.fimd: pm_genpd_resume() [ 77.713912] wake disabled for irq 184 [ 77.713923] wake disabled for irq 185 [ 77.714082] wake disabled for irq 173 [ 77.715676] wake disabled for irq 176 [ 77.718540] exynos4-fb 11c00000.fimd: pm_genpd_runtime_resume() [ 77.718567] exynos4-fb 11c00000.fimd: state restore latency exceeded, new value 1708 ns [ 77.718636] exynos-dsi 11c80000.dsi: pm_genpd_runtime_resume() [ 77.892366] exynos-dsi 11c80000.dsi: PLL failed to stabilize [ 77.892377] exynos-dsi 11c80000.dsi: failed to configure DSI PLL [ 78.192168] exynos-dsi 11c80000.dsi: timeout waiting for reset [ 78.211578] exynos-dsi 11c80000.dsi: waiting for bus lanes timed out [ 78.307173] exynos-dsi 11c80000.dsi: xfer timed out: d1 00 (null) [ 78.307190] panel_s6e8aa0 11c80000.dsi.0: error -110 reading dcs seq(0xd1) [ 78.307199] panel_s6e8aa0 11c80000.dsi.0: read id failed Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> --- drivers/base/power/domain.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)