| Message ID | 1509347446-26105-1-git-send-email-t-kristo@ti.com (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Delegated to: | Rafael Wysocki |
| Headers | show |
On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: > The recent change to the PM QoS framework to introduce a proper > no constraint value overlooked to handle the devices which don't > implement PM QoS OPS. Runtime PM is one of the more severely > impacted subsystems, failing every attempt to runtime suspend > a device. This leads into some nasty second level issues like > probe failures and increased power consumption among other things. Oh, that's bad. Sorry about breaking it and thanks for the fix! > Fix this by adding a proper return value for devices that don't > implement PM QoS implicitly. > > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") > Signed-off-by: Tero Kristo <t-kristo@ti.com> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Applied. > --- > include/linux/pm_qos.h | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h > index 6737a8c..d68b056 100644 > --- a/include/linux/pm_qos.h > +++ b/include/linux/pm_qos.h > @@ -175,7 +175,8 @@ static inline s32 dev_pm_qos_requested_flags(struct device *dev) > static inline s32 dev_pm_qos_raw_read_value(struct device *dev) > { > return IS_ERR_OR_NULL(dev->power.qos) ? > - 0 : pm_qos_read_value(&dev->power.qos->resume_latency); > + PM_QOS_RESUME_LATENCY_NO_CONSTRAINT : > + pm_qos_read_value(&dev->power.qos->resume_latency); > } > #else > static inline enum pm_qos_flags_status __dev_pm_qos_flags(struct device *dev, > -- > 1.9.1 > > -- > Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: > On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: > > The recent change to the PM QoS framework to introduce a proper > > no constraint value overlooked to handle the devices which don't > > implement PM QoS OPS. Runtime PM is one of the more severely > > impacted subsystems, failing every attempt to runtime suspend > > a device. This leads into some nasty second level issues like > > probe failures and increased power consumption among other things. > > Oh, that's bad. > > Sorry about breaking it and thanks for the fix! > > > Fix this by adding a proper return value for devices that don't > > implement PM QoS implicitly. > > > > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") > > Signed-off-by: Tero Kristo <t-kristo@ti.com> > > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > Applied. And pushed to Linus. That said, probe shouldn't ever fail if PM QoS is set to the "never suspend" value. User space can set it that way, after all, so the drivers that fail to probe in that case aren't correct I'm afraid. Thanks, Rafael
On 31/10/17 01:27, Rafael J. Wysocki wrote: > On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>> The recent change to the PM QoS framework to introduce a proper >>> no constraint value overlooked to handle the devices which don't >>> implement PM QoS OPS. Runtime PM is one of the more severely >>> impacted subsystems, failing every attempt to runtime suspend >>> a device. This leads into some nasty second level issues like >>> probe failures and increased power consumption among other things. >> >> Oh, that's bad. >> >> Sorry about breaking it and thanks for the fix! >> >>> Fix this by adding a proper return value for devices that don't >>> implement PM QoS implicitly. >>> >>> Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>> Signed-off-by: Tero Kristo <t-kristo@ti.com> >>> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >> >> Applied. > > And pushed to Linus. > > That said, probe shouldn't ever fail if PM QoS is set to the > "never suspend" value. > > User space can set it that way, after all, so the drivers that fail to probe > in that case aren't correct I'm afraid. Ok interesting. The probe failure we had was a second order issue. A driver (omap_nmailbox) was attempting to pm_runtime_get_sync() ...put_sync() during probe, and checked the return value of pm_runtime_put_sync() which was -EPERM and bailed out. Most of the time, drivers don't check the return value of this and will just succeed. I did a grep on kernel and there are few other drivers that check the return value also, didn't check if they do this during probe though but it can potentially cause various issues elsewhere also. So, you are saying we should not check the return value of pm_runtime_put_x() ever, or should check if it is -EPERM and just pass in that case? Is there any point returning -EPERM from the runtime core at all then? This should probably be filtered out within runtime core as a valid situation and just return 0. -Tero -- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On Tue, Oct 31, 2017 at 8:13 AM, Tero Kristo <t-kristo@ti.com> wrote: > On 31/10/17 01:27, Rafael J. Wysocki wrote: >> >> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >>> >>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>>> >>>> The recent change to the PM QoS framework to introduce a proper >>>> no constraint value overlooked to handle the devices which don't >>>> implement PM QoS OPS. Runtime PM is one of the more severely >>>> impacted subsystems, failing every attempt to runtime suspend >>>> a device. This leads into some nasty second level issues like >>>> probe failures and increased power consumption among other things. >>> >>> >>> Oh, that's bad. >>> >>> Sorry about breaking it and thanks for the fix! >>> >>>> Fix this by adding a proper return value for devices that don't >>>> implement PM QoS implicitly. >>>> >>>> Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>>> Signed-off-by: Tero Kristo <t-kristo@ti.com> >>>> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >>> >>> >>> Applied. >> >> >> And pushed to Linus. >> >> That said, probe shouldn't ever fail if PM QoS is set to the >> "never suspend" value. >> >> User space can set it that way, after all, so the drivers that fail to >> probe >> in that case aren't correct I'm afraid. > > > Ok interesting. The probe failure we had was a second order issue. A driver > (omap_nmailbox) was attempting to pm_runtime_get_sync() ...put_sync() during > probe, and checked the return value of pm_runtime_put_sync() which was > -EPERM and bailed out. Most of the time, drivers don't check the return > value of this and will just succeed. I did a grep on kernel and there are > few other drivers that check the return value also, didn't check if they do > this during probe though but it can potentially cause various issues > elsewhere also. > > So, you are saying we should not check the return value of > pm_runtime_put_x() ever, or should check if it is -EPERM and just pass in > that case? The latter. > Is there any point returning -EPERM from the runtime core at all > then? This should probably be filtered out within runtime core as a valid > situation and just return 0. Fair point. However, there are other situations in which pm_runtime_put_sync() can return an error code which needs to be checked, like -EBUSY or -EAGAIN returned if one of the reference counters is nonzero. In fact, the "no suspend" PM QoS constraint is somewhat similar to this situation, so what about changing the error code returned to -EAGAIN, for example?
On 31/10/17 10:40, Rafael J. Wysocki wrote: > On Tue, Oct 31, 2017 at 8:13 AM, Tero Kristo <t-kristo@ti.com> wrote: >> On 31/10/17 01:27, Rafael J. Wysocki wrote: >>> >>> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >>>> >>>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>>>> >>>>> The recent change to the PM QoS framework to introduce a proper >>>>> no constraint value overlooked to handle the devices which don't >>>>> implement PM QoS OPS. Runtime PM is one of the more severely >>>>> impacted subsystems, failing every attempt to runtime suspend >>>>> a device. This leads into some nasty second level issues like >>>>> probe failures and increased power consumption among other things. >>>> >>>> >>>> Oh, that's bad. >>>> >>>> Sorry about breaking it and thanks for the fix! >>>> >>>>> Fix this by adding a proper return value for devices that don't >>>>> implement PM QoS implicitly. >>>>> >>>>> Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>>>> Signed-off-by: Tero Kristo <t-kristo@ti.com> >>>>> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >>>> >>>> >>>> Applied. >>> >>> >>> And pushed to Linus. >>> >>> That said, probe shouldn't ever fail if PM QoS is set to the >>> "never suspend" value. >>> >>> User space can set it that way, after all, so the drivers that fail to >>> probe >>> in that case aren't correct I'm afraid. >> >> >> Ok interesting. The probe failure we had was a second order issue. A driver >> (omap_nmailbox) was attempting to pm_runtime_get_sync() ...put_sync() during >> probe, and checked the return value of pm_runtime_put_sync() which was >> -EPERM and bailed out. Most of the time, drivers don't check the return >> value of this and will just succeed. I did a grep on kernel and there are >> few other drivers that check the return value also, didn't check if they do >> this during probe though but it can potentially cause various issues >> elsewhere also. >> >> So, you are saying we should not check the return value of >> pm_runtime_put_x() ever, or should check if it is -EPERM and just pass in >> that case? > > The latter. > >> Is there any point returning -EPERM from the runtime core at all >> then? This should probably be filtered out within runtime core as a valid >> situation and just return 0. > > Fair point. > > However, there are other situations in which pm_runtime_put_sync() can > return an error code which needs to be checked, like -EBUSY or -EAGAIN > returned if one of the reference counters is nonzero. > > In fact, the "no suspend" PM QoS constraint is somewhat similar to > this situation, so what about changing the error code returned to > -EAGAIN, for example? Ok yea thats true. Its better to leave the decision to the drivers as the core most likely can't know what the driver will actually want to do with a prevented pm_runtime_put. That way the policy can be implemented on case-by-case basis. Keeping the -EPERM in place is probably also just fine, if someone needs a detailed info of what prevented the pm_runtime_put from finishing properly, and if someone is using this value already. -Tero -- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
Hi Rafael, Tero, On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote: > On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >> > The recent change to the PM QoS framework to introduce a proper >> > no constraint value overlooked to handle the devices which don't >> > implement PM QoS OPS. Runtime PM is one of the more severely >> > impacted subsystems, failing every attempt to runtime suspend >> > a device. This leads into some nasty second level issues like >> > probe failures and increased power consumption among other things. >> >> Oh, that's bad. >> >> Sorry about breaking it and thanks for the fix! >> >> > Fix this by adding a proper return value for devices that don't >> > implement PM QoS implicitly. >> > >> > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >> > Signed-off-by: Tero Kristo <t-kristo@ti.com> >> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >> >> Applied. > > And pushed to Linus. I'm afraid it is not sufficient. Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") introduced two issues on Renesas platforms: 1. After boot up, many devices have changed their state from "suspended" to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary (comparing that file across boots is one of my standard tests). Interestingly, doing a system suspend/resume cycle restores their state to "suspended". 2. During system suspend, the following warning is printed on r8a7791/koelsch: i2c-rcar e6530000.i2c: runtime PM trying to suspend device but active child Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume latency") fixes the second issue, but not the first. Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") fixes both. Do you have a clue? Thanks! 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
CC linux-renesas-soc On Tue, Oct 31, 2017 at 2:09 PM, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > Hi Rafael, Tero, > > On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote: >> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>> > The recent change to the PM QoS framework to introduce a proper >>> > no constraint value overlooked to handle the devices which don't >>> > implement PM QoS OPS. Runtime PM is one of the more severely >>> > impacted subsystems, failing every attempt to runtime suspend >>> > a device. This leads into some nasty second level issues like >>> > probe failures and increased power consumption among other things. >>> >>> Oh, that's bad. >>> >>> Sorry about breaking it and thanks for the fix! >>> >>> > Fix this by adding a proper return value for devices that don't >>> > implement PM QoS implicitly. >>> > >>> > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>> > Signed-off-by: Tero Kristo <t-kristo@ti.com> >>> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >>> >>> Applied. >> >> And pushed to Linus. > > I'm afraid it is not sufficient. > > Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") > introduced two issues on Renesas platforms: > 1. After boot up, many devices have changed their state from "suspended" > to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary > (comparing that file across boots is one of my standard tests). > Interestingly, doing a system suspend/resume cycle restores their state > to "suspended". > > 2. During system suspend, the following warning is printed on > r8a7791/koelsch: > > i2c-rcar e6530000.i2c: runtime PM trying to suspend device but > active child > > Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume > latency") fixes the second issue, but not the first. > > Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm > device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume > latency PM QoS") fixes both. > > Do you have a clue? > Thanks!
Hi Rafael, Tero, CC pinchartl, dri-devel On Tue, Oct 31, 2017 at 2:10 PM, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > CC linux-renesas-soc > > On Tue, Oct 31, 2017 at 2:09 PM, Geert Uytterhoeven > <geert@linux-m68k.org> wrote: >> On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote: >>> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >>>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>>> > The recent change to the PM QoS framework to introduce a proper >>>> > no constraint value overlooked to handle the devices which don't >>>> > implement PM QoS OPS. Runtime PM is one of the more severely >>>> > impacted subsystems, failing every attempt to runtime suspend >>>> > a device. This leads into some nasty second level issues like >>>> > probe failures and increased power consumption among other things. >>>> >>>> Oh, that's bad. >>>> >>>> Sorry about breaking it and thanks for the fix! >>>> >>>> > Fix this by adding a proper return value for devices that don't >>>> > implement PM QoS implicitly. >>>> > >>>> > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>>> > Signed-off-by: Tero Kristo <t-kristo@ti.com> >>>> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >>>> >>>> Applied. >>> >>> And pushed to Linus. >> >> I'm afraid it is not sufficient. >> >> Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") >> introduced two issues on Renesas platforms: >> 1. After boot up, many devices have changed their state from "suspended" >> to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary >> (comparing that file across boots is one of my standard tests). >> Interestingly, doing a system suspend/resume cycle restores their state >> to "suspended". >> >> 2. During system suspend, the following warning is printed on >> r8a7791/koelsch: >> >> i2c-rcar e6530000.i2c: runtime PM trying to suspend device but >> active child 3. I've just bisected a seemingly unrelated issue to the same commit. On Salvator-XS with R-Car H3, initialization of the rcar-du driver now takes more than 1 minute due to flip_done time outs, while it took 0.12s before: [ 3.015035] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013). [ 3.021721] [drm] No driver support for vblank timestamp query. [ 13.280738] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* [CRTC:58:crtc-3] flip_done timed out [ 23.520707] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* [CRTC:58:crtc-3] flip_done timed out [ 33.760708] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* [CRTC:58:crtc-3] flip_done timed out [ 44.000755] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* [CRTC:58:crtc-3] flip_done timed out [ 44.003597] Console: switching to colour frame buffer device 128x48 [ 54.240707] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* [CRTC:58:crtc-3] flip_done timed out [ 64.480706] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* [CRTC:58:crtc-3] flip_done timed out [ 64.544876] rcar-du feb00000.display: fb0: frame buffer device [ 64.552013] [drm] Initialized rcar-du 1.0.0 20130110 for feb00000.display on minor 0 [ 64.559873] [drm] Device feb00000.display probed >> Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume >> latency") fixes the second issue, but not the first. ... nor the third. >> Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm >> device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume >> latency PM QoS") fixes both. ... all three. >> Do you have a clue? >> Thanks! 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
On 31 October 2017 at 14:55, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > Hi Rafael, Tero, > > CC pinchartl, dri-devel > > On Tue, Oct 31, 2017 at 2:10 PM, Geert Uytterhoeven > <geert@linux-m68k.org> wrote: >> CC linux-renesas-soc >> >> On Tue, Oct 31, 2017 at 2:09 PM, Geert Uytterhoeven >> <geert@linux-m68k.org> wrote: >>> On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote: >>>> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >>>>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>>>> > The recent change to the PM QoS framework to introduce a proper >>>>> > no constraint value overlooked to handle the devices which don't >>>>> > implement PM QoS OPS. Runtime PM is one of the more severely >>>>> > impacted subsystems, failing every attempt to runtime suspend >>>>> > a device. This leads into some nasty second level issues like >>>>> > probe failures and increased power consumption among other things. >>>>> >>>>> Oh, that's bad. >>>>> >>>>> Sorry about breaking it and thanks for the fix! >>>>> >>>>> > Fix this by adding a proper return value for devices that don't >>>>> > implement PM QoS implicitly. >>>>> > >>>>> > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>>>> > Signed-off-by: Tero Kristo <t-kristo@ti.com> >>>>> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >>>>> >>>>> Applied. >>>> >>>> And pushed to Linus. >>> >>> I'm afraid it is not sufficient. >>> >>> Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") >>> introduced two issues on Renesas platforms: >>> 1. After boot up, many devices have changed their state from "suspended" >>> to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary >>> (comparing that file across boots is one of my standard tests). >>> Interestingly, doing a system suspend/resume cycle restores their state >>> to "suspended". >>> >>> 2. During system suspend, the following warning is printed on >>> r8a7791/koelsch: >>> >>> i2c-rcar e6530000.i2c: runtime PM trying to suspend device but >>> active child > > 3. I've just bisected a seemingly unrelated issue to the same commit. > On Salvator-XS with R-Car H3, initialization of the rcar-du driver now > takes more than 1 minute due to flip_done time outs, while it took 0.12s > before: > > [ 3.015035] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013). > [ 3.021721] [drm] No driver support for vblank timestamp query. > [ 13.280738] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 23.520707] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 33.760708] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 44.000755] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 44.003597] Console: switching to colour frame buffer device 128x48 > [ 54.240707] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 64.480706] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 64.544876] rcar-du feb00000.display: fb0: frame buffer device > [ 64.552013] [drm] Initialized rcar-du 1.0.0 20130110 for > feb00000.display on minor 0 > [ 64.559873] [drm] Device feb00000.display probed > >>> Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume >>> latency") fixes the second issue, but not the first. > > ... nor the third. > >>> Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm >>> device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume >>> latency PM QoS") fixes both. > > ... all three. > >>> Do you have a clue? >>> Thanks! As I didn't have the time to review the original commit, before it got pushed as a fix, I am planning to review it now instead. A vague guess is that the genpd governor prevents the device from being suspended. That was also the most tricky part of the changes from the original commit, which is causing problems. I get back to this when I have reviewed it more thoroughly. Kind regards Uffe
On Tue, 31 Oct 2017, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > Hi Rafael, Tero, > > CC pinchartl, dri-devel Cc: Marta, Martin Our CI is hitting this too. BR, Jani. > > On Tue, Oct 31, 2017 at 2:10 PM, Geert Uytterhoeven > <geert@linux-m68k.org> wrote: >> CC linux-renesas-soc >> >> On Tue, Oct 31, 2017 at 2:09 PM, Geert Uytterhoeven >> <geert@linux-m68k.org> wrote: >>> On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote: >>>> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >>>>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>>>> > The recent change to the PM QoS framework to introduce a proper >>>>> > no constraint value overlooked to handle the devices which don't >>>>> > implement PM QoS OPS. Runtime PM is one of the more severely >>>>> > impacted subsystems, failing every attempt to runtime suspend >>>>> > a device. This leads into some nasty second level issues like >>>>> > probe failures and increased power consumption among other things. >>>>> >>>>> Oh, that's bad. >>>>> >>>>> Sorry about breaking it and thanks for the fix! >>>>> >>>>> > Fix this by adding a proper return value for devices that don't >>>>> > implement PM QoS implicitly. >>>>> > >>>>> > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>>>> > Signed-off-by: Tero Kristo <t-kristo@ti.com> >>>>> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >>>>> >>>>> Applied. >>>> >>>> And pushed to Linus. >>> >>> I'm afraid it is not sufficient. >>> >>> Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") >>> introduced two issues on Renesas platforms: >>> 1. After boot up, many devices have changed their state from "suspended" >>> to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary >>> (comparing that file across boots is one of my standard tests). >>> Interestingly, doing a system suspend/resume cycle restores their state >>> to "suspended". >>> >>> 2. During system suspend, the following warning is printed on >>> r8a7791/koelsch: >>> >>> i2c-rcar e6530000.i2c: runtime PM trying to suspend device but >>> active child > > 3. I've just bisected a seemingly unrelated issue to the same commit. > On Salvator-XS with R-Car H3, initialization of the rcar-du driver now > takes more than 1 minute due to flip_done time outs, while it took 0.12s > before: > > [ 3.015035] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013). > [ 3.021721] [drm] No driver support for vblank timestamp query. > [ 13.280738] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 23.520707] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 33.760708] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 44.000755] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 44.003597] Console: switching to colour frame buffer device 128x48 > [ 54.240707] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 64.480706] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 64.544876] rcar-du feb00000.display: fb0: frame buffer device > [ 64.552013] [drm] Initialized rcar-du 1.0.0 20130110 for > feb00000.display on minor 0 > [ 64.559873] [drm] Device feb00000.display probed > >>> Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume >>> latency") fixes the second issue, but not the first. > > ... nor the third. > >>> Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm >>> device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume >>> latency PM QoS") fixes both. > > ... all three. > >>> Do you have a clue? >>> Thanks!
On Tue, Oct 31, 2017 at 3:04 PM, Ulf Hansson <ulf.hansson@linaro.org> wrote: > On 31 October 2017 at 14:55, Geert Uytterhoeven <geert@linux-m68k.org> wrote: >> Hi Rafael, Tero, >> >> CC pinchartl, dri-devel >> >> On Tue, Oct 31, 2017 at 2:10 PM, Geert Uytterhoeven >> <geert@linux-m68k.org> wrote: >>> CC linux-renesas-soc >>> >>> On Tue, Oct 31, 2017 at 2:09 PM, Geert Uytterhoeven >>> <geert@linux-m68k.org> wrote: >>>> On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote: >>>>> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: >>>>>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: >>>>>> > The recent change to the PM QoS framework to introduce a proper >>>>>> > no constraint value overlooked to handle the devices which don't >>>>>> > implement PM QoS OPS. Runtime PM is one of the more severely >>>>>> > impacted subsystems, failing every attempt to runtime suspend >>>>>> > a device. This leads into some nasty second level issues like >>>>>> > probe failures and increased power consumption among other things. >>>>>> >>>>>> Oh, that's bad. >>>>>> >>>>>> Sorry about breaking it and thanks for the fix! >>>>>> >>>>>> > Fix this by adding a proper return value for devices that don't >>>>>> > implement PM QoS implicitly. >>>>>> > >>>>>> > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") >>>>>> > Signed-off-by: Tero Kristo <t-kristo@ti.com> >>>>>> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> >>>>>> >>>>>> Applied. >>>>> >>>>> And pushed to Linus. >>>> >>>> I'm afraid it is not sufficient. >>>> >>>> Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") >>>> introduced two issues on Renesas platforms: >>>> 1. After boot up, many devices have changed their state from "suspended" >>>> to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary >>>> (comparing that file across boots is one of my standard tests). >>>> Interestingly, doing a system suspend/resume cycle restores their state >>>> to "suspended". >>>> >>>> 2. During system suspend, the following warning is printed on >>>> r8a7791/koelsch: >>>> >>>> i2c-rcar e6530000.i2c: runtime PM trying to suspend device but >>>> active child >> >> 3. I've just bisected a seemingly unrelated issue to the same commit. >> On Salvator-XS with R-Car H3, initialization of the rcar-du driver now >> takes more than 1 minute due to flip_done time outs, while it took 0.12s >> before: >> >> [ 3.015035] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013). >> [ 3.021721] [drm] No driver support for vblank timestamp query. >> [ 13.280738] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* >> [CRTC:58:crtc-3] flip_done timed out >> [ 23.520707] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* >> [CRTC:58:crtc-3] flip_done timed out >> [ 33.760708] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* >> [CRTC:58:crtc-3] flip_done timed out >> [ 44.000755] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* >> [CRTC:58:crtc-3] flip_done timed out >> [ 44.003597] Console: switching to colour frame buffer device 128x48 >> [ 54.240707] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* >> [CRTC:58:crtc-3] flip_done timed out >> [ 64.480706] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* >> [CRTC:58:crtc-3] flip_done timed out >> [ 64.544876] rcar-du feb00000.display: fb0: frame buffer device >> [ 64.552013] [drm] Initialized rcar-du 1.0.0 20130110 for >> feb00000.display on minor 0 >> [ 64.559873] [drm] Device feb00000.display probed >> >>>> Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume >>>> latency") fixes the second issue, but not the first. >> >> ... nor the third. >> >>>> Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm >>>> device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume >>>> latency PM QoS") fixes both. >> >> ... all three. >> >>>> Do you have a clue? >>>> Thanks! > > As I didn't have the time to review the original commit, before it got > pushed as a fix, I am planning to review it now instead. > > A vague guess is that the genpd governor prevents the device from > being suspended. That was also the most tricky part of the changes > from the original commit, which is causing problems. I think you are right. > I get back to this when I have reviewed it more thoroughly. Thanks, and sorry for breaking stuff. In retrospect I should just have not pushed it so late in the cycle. Rafael
On Tue, Oct 31, 2017 at 05:37:50PM +0200, Jani Nikula wrote: > On Tue, 31 Oct 2017, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > > Hi Rafael, Tero, > > > > CC pinchartl, dri-devel > > Cc: Marta, Martin > > Our CI is hitting this too. Should be ok again, we've locally reverted this patch. But a big chunk of the pw runs overnight did hit it unfortunately :-( -Daniel > > BR, > Jani. > > > > > On Tue, Oct 31, 2017 at 2:10 PM, Geert Uytterhoeven > > <geert@linux-m68k.org> wrote: > >> CC linux-renesas-soc > >> > >> On Tue, Oct 31, 2017 at 2:09 PM, Geert Uytterhoeven > >> <geert@linux-m68k.org> wrote: > >>> On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote: > >>>> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: > >>>>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: > >>>>> > The recent change to the PM QoS framework to introduce a proper > >>>>> > no constraint value overlooked to handle the devices which don't > >>>>> > implement PM QoS OPS. Runtime PM is one of the more severely > >>>>> > impacted subsystems, failing every attempt to runtime suspend > >>>>> > a device. This leads into some nasty second level issues like > >>>>> > probe failures and increased power consumption among other things. > >>>>> > >>>>> Oh, that's bad. > >>>>> > >>>>> Sorry about breaking it and thanks for the fix! > >>>>> > >>>>> > Fix this by adding a proper return value for devices that don't > >>>>> > implement PM QoS implicitly. > >>>>> > > >>>>> > Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") > >>>>> > Signed-off-by: Tero Kristo <t-kristo@ti.com> > >>>>> > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > >>>>> > >>>>> Applied. > >>>> > >>>> And pushed to Linus. > >>> > >>> I'm afraid it is not sufficient. > >>> > >>> Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") > >>> introduced two issues on Renesas platforms: > >>> 1. After boot up, many devices have changed their state from "suspended" > >>> to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary > >>> (comparing that file across boots is one of my standard tests). > >>> Interestingly, doing a system suspend/resume cycle restores their state > >>> to "suspended". > >>> > >>> 2. During system suspend, the following warning is printed on > >>> r8a7791/koelsch: > >>> > >>> i2c-rcar e6530000.i2c: runtime PM trying to suspend device but > >>> active child > > > > 3. I've just bisected a seemingly unrelated issue to the same commit. > > On Salvator-XS with R-Car H3, initialization of the rcar-du driver now > > takes more than 1 minute due to flip_done time outs, while it took 0.12s > > before: > > > > [ 3.015035] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013). > > [ 3.021721] [drm] No driver support for vblank timestamp query. > > [ 13.280738] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > > [CRTC:58:crtc-3] flip_done timed out > > [ 23.520707] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > > [CRTC:58:crtc-3] flip_done timed out > > [ 33.760708] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > > [CRTC:58:crtc-3] flip_done timed out > > [ 44.000755] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > > [CRTC:58:crtc-3] flip_done timed out > > [ 44.003597] Console: switching to colour frame buffer device 128x48 > > [ 54.240707] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > > [CRTC:58:crtc-3] flip_done timed out > > [ 64.480706] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > > [CRTC:58:crtc-3] flip_done timed out > > [ 64.544876] rcar-du feb00000.display: fb0: frame buffer device > > [ 64.552013] [drm] Initialized rcar-du 1.0.0 20130110 for > > feb00000.display on minor 0 > > [ 64.559873] [drm] Device feb00000.display probed > > > >>> Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume > >>> latency") fixes the second issue, but not the first. > > > > ... nor the third. > > > >>> Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm > >>> device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume > >>> latency PM QoS") fixes both. > > > > ... all three. > > > >>> Do you have a clue? > >>> Thanks! > > > -- > Jani Nikula, Intel Open Source Technology Center > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/dri-devel
Hi Geert, On Tuesday, 31 October 2017 15:55:02 EET Geert Uytterhoeven wrote: > On Tue, Oct 31, 2017 at 2:10 PM, Geert Uytterhoeven wrote: > > On Tue, Oct 31, 2017 at 2:09 PM, Geert Uytterhoeven wrote: > >> On Tue, Oct 31, 2017 at 12:27 AM, Rafael J. Wysocki wrote: > >>> On Monday, October 30, 2017 11:19:08 AM CET Rafael J. Wysocki wrote: > >>>> On Mon, Oct 30, 2017 at 8:10 AM, Tero Kristo <t-kristo@ti.com> wrote: > >>>>> The recent change to the PM QoS framework to introduce a proper > >>>>> no constraint value overlooked to handle the devices which don't > >>>>> implement PM QoS OPS. Runtime PM is one of the more severely > >>>>> impacted subsystems, failing every attempt to runtime suspend > >>>>> a device. This leads into some nasty second level issues like > >>>>> probe failures and increased power consumption among other things. > >>>> > >>>> Oh, that's bad. > >>>> > >>>> Sorry about breaking it and thanks for the fix! > >>>> > >>>>> Fix this by adding a proper return value for devices that don't > >>>>> implement PM QoS implicitly. > >>>>> > >>>>> Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") > >>>>> Signed-off-by: Tero Kristo <t-kristo@ti.com> > >>>>> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > >>>> > >>>> Applied. > >>> > >>> And pushed to Linus. > >> > >> I'm afraid it is not sufficient. > >> > >> Commit 0cc2b4e5a020fc7f ("PM / QoS: Fix device resume latency PM QoS") > >> > >> introduced two issues on Renesas platforms: > >> 1. After boot up, many devices have changed their state from "suspended" > >> to "active", according to /sys/kernel/debug/pm_genpd/pm_genpd_summary > >> (comparing that file across boots is one of my standard tests). > >> Interestingly, doing a system suspend/resume cycle restores their > >> state to "suspended". > >> > >> 2. During system suspend, the following warning is printed on > >> r8a7791/koelsch: > >> i2c-rcar e6530000.i2c: runtime PM trying to suspend device but > >> > >> active child > > 3. I've just bisected a seemingly unrelated issue to the same commit. > On Salvator-XS with R-Car H3, initialization of the rcar-du driver now > takes more than 1 minute due to flip_done time outs, while it took 0.12s > before: > > [ 3.015035] [drm] Supports vblank timestamp caching Rev 2 > (21.10.2013). [ 3.021721] [drm] No driver support for vblank timestamp > query. [ 13.280738] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 23.520707] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 33.760708] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 44.000755] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 44.003597] Console: switching to colour frame buffer device 128x48 > [ 54.240707] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 64.480706] [drm:drm_atomic_helper_commit_cleanup_done] *ERROR* > [CRTC:58:crtc-3] flip_done timed out > [ 64.544876] rcar-du feb00000.display: fb0: frame buffer device > [ 64.552013] [drm] Initialized rcar-du 1.0.0 20130110 for > feb00000.display on minor 0 > [ 64.559873] [drm] Device feb00000.display probed > > >> Commit 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm device resume > >> latency") fixes the second issue, but not the first. > > ... nor the third. > > >> Reverting commits 2a9a86d5c81389cd ("PM / QoS: Fix default runtime_pm > >> device resume latency") and 0cc2b4e5a020fc7f ("PM / QoS: Fix device > >> resume > >> latency PM QoS") fixes both. > > ... all three. Thank you for tracking this and notifying me. I like it even better now that the problem seems to be fixed without requiring any action from my side :-) > >> Do you have a clue? > >> Thanks!
diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h index 6737a8c..d68b056 100644 --- a/include/linux/pm_qos.h +++ b/include/linux/pm_qos.h @@ -175,7 +175,8 @@ static inline s32 dev_pm_qos_requested_flags(struct device *dev) static inline s32 dev_pm_qos_raw_read_value(struct device *dev) { return IS_ERR_OR_NULL(dev->power.qos) ? - 0 : pm_qos_read_value(&dev->power.qos->resume_latency); + PM_QOS_RESUME_LATENCY_NO_CONSTRAINT : + pm_qos_read_value(&dev->power.qos->resume_latency); } #else static inline enum pm_qos_flags_status __dev_pm_qos_flags(struct device *dev,
The recent change to the PM QoS framework to introduce a proper no constraint value overlooked to handle the devices which don't implement PM QoS OPS. Runtime PM is one of the more severely impacted subsystems, failing every attempt to runtime suspend a device. This leads into some nasty second level issues like probe failures and increased power consumption among other things. Fix this by adding a proper return value for devices that don't implement PM QoS implicitly. Fixes: 0cc2b4e5a020 ("PM / QoS: Fix device resume latency PM QoS") Signed-off-by: Tero Kristo <t-kristo@ti.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> --- include/linux/pm_qos.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)