DC6 already programmed to be disabled

DC6 already programmed to be disabled

[Konduru, Chandra]

[-- Attachment #1.1: Type: text/plain, Size: 4229 bytes --]

Hi,
I have been seeing below warning on skylake system on which dmc fw isn't placed.
Is below warning expected? If so what is it conveying?

-Chandra

May 15 15:53:53 cmkondur-skl kernel: [19587.475013] WARNING: CPU: 0 PID: 1453 at /home/cmkondur/otc/otc_kernel/drivers/gpu/drm/i915/intel_runtime_pm.c:532 skl_set_power_well+0x74b/0xab0 [i915]()
May 15 15:53:53 cmkondur-skl kernel: [19587.475014] DC6 already programmed to be disabled.
May 15 15:53:53 cmkondur-skl kernel: [19587.475015] Modules linked in: i915(OE) drm_kms_helper(OE) drm(OE) video(E) i2c_algo_bit(E) xhci_plat_hcd(E) snd_hda_intel(E) snd_hda_codec(E) snd_hda_core(E) snd_hwdep(E) snd_pcm_oss(E) snd_mixer_oss(E) snd_pcm(E) dwc3(E) udc_core(E) snd_seq_midi(E) snd_rawmidi(E) intel_rapl(E) snd_seq_midi_event(E) iosf_mbi(E) x86_pkg_temp_thermal(E) coretemp(E) snd_seq(E) kvm_intel(E) kvm(E) crct10dif_pclmul(E) crc32_pclmul(E) rfcomm(E) bnep(E) ghash_clmulni_intel(E) bluetooth(E) snd_timer(E) snd_seq_device(E) aesni_intel(E) ablk_helper(E) cryptd(E) lrw(E) gf128mul(E) glue_helper(E) aes_x86_64(E) asix(E) usbnet(E) mii(E) snd(E) mac_hid(E) serio_raw(E) nls_iso8859_1(E) soundcore(E) dwc3_pci(E) parport_pc(E) ppdev(E) lp(E) parport(E) hid_generic(E) usbhid(E) hid(E) e1000e(E) ptp(E) ahci(E) psmouse(E) pps_core(E) sdhci_pci(E) sdhci(E) libahci(E)
May 15 15:53:53 cmkondur-skl kernel: [19587.475041] CPU: 0 PID: 1453 Comm: kms_nv12 Tainted: G        W  OE   4.1.0-rc2+ #2
May 15 15:53:53 cmkondur-skl kernel: [19587.475042] Hardware name: Intel Corporation Skylake Client platform/Skylake Y LPDDR3 RVP3, BIOS SKLSE2R1.86C.X075.R04.1503240732 03/24/2015
May 15 15:53:53 cmkondur-skl kernel: [19587.475043]  0000000000000214 ffff880075547a68 ffffffff817dc684 0000000000000007
May 15 15:53:53 cmkondur-skl kernel: [19587.475045]  ffff880075547ab8 ffff880075547aa8 ffffffff8107c557 000000003000000f
May 15 15:53:53 cmkondur-skl kernel: [19587.475047]  ffff8800755e0000 0000000040000000 ffffffffc05897f0 000000003000000f
May 15 15:53:53 cmkondur-skl kernel: [19587.475048] Call Trace:
May 15 15:53:53 cmkondur-skl kernel: [19587.475050]  [<ffffffff817dc684>] dump_stack+0x45/0x57
May 15 15:53:53 cmkondur-skl kernel: [19587.475052]  [<ffffffff8107c557>] warn_slowpath_common+0x97/0xe0
May 15 15:53:53 cmkondur-skl kernel: [19587.475054]  [<ffffffff8107c656>] warn_slowpath_fmt+0x46/0x50
May 15 15:53:53 cmkondur-skl kernel: [19587.475062]  [<ffffffffc04ad40b>] skl_set_power_well+0x74b/0xab0 [i915]
May 15 15:53:53 cmkondur-skl kernel: [19587.475071]  [<ffffffffc04ad7a3>] skl_power_well_enable+0x13/0x20 [i915]
May 15 15:53:53 cmkondur-skl kernel: [19587.475079]  [<ffffffffc04ada6e>] intel_display_power_get+0x9e/0xf0 [i915]
May 15 15:53:53 cmkondur-skl kernel: [19587.475095]  [<ffffffffc0536c51>] intel_dp_detect+0xb1/0x510 [i915]
May 15 15:53:53 cmkondur-skl kernel: [19587.475099]  [<ffffffffc047dc10>] drm_helper_probe_single_connector_modes_merge_bits+0x200/0x4f0 [drm_kms_helper]
May 15 15:53:53 cmkondur-skl kernel: [19587.475108]  [<ffffffffc0400953>] ? _object_find+0x93/0xd0 [drm]
May 15 15:53:53 cmkondur-skl kernel: [19587.475112]  [<ffffffffc047df33>] drm_helper_probe_single_connector_modes+0x13/0x20 [drm_kms_helper]
May 15 15:53:53 cmkondur-skl kernel: [19587.475121]  [<ffffffffc0404b9f>] drm_mode_getconnector+0x3af/0x400 [drm]
May 15 15:53:53 cmkondur-skl kernel: [19587.475124]  [<ffffffff817d8350>] ? printk+0x46/0x48
May 15 15:53:53 cmkondur-skl kernel: [19587.475130]  [<ffffffffc03f5cb2>] drm_ioctl+0x362/0x690 [drm]
May 15 15:53:53 cmkondur-skl kernel: [19587.475139]  [<ffffffffc04047f0>] ? drm_mode_getcrtc+0x150/0x150 [drm]
May 15 15:53:53 cmkondur-skl kernel: [19587.475141]  [<ffffffff81211406>] do_vfs_ioctl+0x86/0x530
May 15 15:53:53 cmkondur-skl kernel: [19587.475144]  [<ffffffff8121e7c4>] ? mntput+0x24/0x40
May 15 15:53:53 cmkondur-skl kernel: [19587.475146]  [<ffffffff811ffe80>] ? __fput+0x170/0x250
May 15 15:53:53 cmkondur-skl kernel: [19587.475148]  [<ffffffff81211941>] SyS_ioctl+0x91/0xb0
May 15 15:53:53 cmkondur-skl kernel: [19587.475150]  [<ffffffff817e4672>] system_call_fastpath+0x16/0x75
May 15 15:53:53 cmkondur-skl kernel: [19587.475151] ---[ end trace cd9e938b57041455 ]---

[-- Attachment #1.2: Type: text/html, Size: 7407 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: DC6 already programmed to be disabled

[Daniel Vetter]

On Fri, May 15, 2015 at 11:22:27PM +0000, Konduru, Chandra wrote:
> Hi,
> I have been seeing below warning on skylake system on which dmc fw isn't placed.
> Is below warning expected? If so what is it conveying?

Seems to be another fallout from the current design of how we prevent
dc5/6 when the firmware is not (yet) loaded. I've detailed how this should
be fixed. We need to prevent the rpm code from ever trying to shut down
that specific power well instead of just not obeying the request. Not
obeying the request means the rpm code is out of sync with reality,
leading to WARN_ON fun like the one you've hit here.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: DC6 already programmed to be disabled

[Damien Lespiau]

On Mon, May 18, 2015 at 10:38:03AM +0200, Daniel Vetter wrote:
> On Fri, May 15, 2015 at 11:22:27PM +0000, Konduru, Chandra wrote:
> > Hi,
> > I have been seeing below warning on skylake system on which dmc fw isn't placed.
> > Is below warning expected? If so what is it conveying?
> 
> Seems to be another fallout from the current design of how we prevent
> dc5/6 when the firmware is not (yet) loaded. I've detailed how this should
> be fixed. We need to prevent the rpm code from ever trying to shut down
> that specific power well instead of just not obeying the request. Not
> obeying the request means the rpm code is out of sync with reality,
> leading to WARN_ON fun like the one you've hit here.

Hey all,

Would anyone of you (Sunil, Animesh, Suketu) have time to fix this? (the
warning when DMC firmware isn't there). We should be able to work when
failing to load the DMC firmware.

What Daniel says is not quite accurate, bear in mind we still can shut
down all power wells and do PC10 with screens off when the DMC isn't
loaded. We could also decide to disable run-time PM entirely when the
DMC firmware isn't there. That's something that can be fixed later on
though, right now the most immediate issue is not to dump lots of
warnings when failing to load the firmware.

By default I'll fix it, I have this on my TODO list, it's quite low
though.

Thanks,

-- 
Damien
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: DC6 already programmed to be disabled

[Kamath, Sunil]

Sure Damien. We will come back with solution for the same. 

- Sunil

>-----Original Message-----
>From: Lespiau, Damien 
>Sent: Monday, May 18, 2015 4:14 PM
>To: Daniel Vetter
>Cc: Konduru, Chandra; intel-gfx@lists.freedesktop.org; Shah, Suketu J; Kamath, Sunil; Manna, Animesh
>Subject: Re: [Intel-gfx] DC6 already programmed to be disabled
>
>On Mon, May 18, 2015 at 10:38:03AM +0200, Daniel Vetter wrote:
>> On Fri, May 15, 2015 at 11:22:27PM +0000, Konduru, Chandra wrote:
>> > Hi,
>> > I have been seeing below warning on skylake system on which dmc fw isn't placed.
>> > Is below warning expected? If so what is it conveying?
>> 
>> Seems to be another fallout from the current design of how we prevent
>> dc5/6 when the firmware is not (yet) loaded. I've detailed how this 
>> should be fixed. We need to prevent the rpm code from ever trying to 
>> shut down that specific power well instead of just not obeying the 
>> request. Not obeying the request means the rpm code is out of sync 
>> with reality, leading to WARN_ON fun like the one you've hit here.
>
>Hey all,
>
>Would anyone of you (Sunil, Animesh, Suketu) have time to fix this? (the warning when DMC firmware isn't there). We should be able to work when failing to load the DMC firmware.
>
>What Daniel says is not quite accurate, bear in mind we still can shut down all power wells and do PC10 with screens off when the DMC isn't loaded. We could also decide to disable run-time PM entirely when the >DMC firmware isn't there. That's something that can be fixed later on though, right now the most immediate issue is not to dump lots of warnings when failing to load the firmware.
>
>By default I'll fix it, I have this on my TODO list, it's quite low though.
>
>Thanks,
>
>--
>Damien
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: DC6 already programmed to be disabled

[Daniel Vetter]

On Tue, May 19, 2015 at 04:53:16AM +0000, Kamath, Sunil wrote:
> Sure Damien. We will come back with solution for the same. 

Please check out my reply to Animesh' patch in

https://tango.freedesktop.org/patch/49084/

I think this is the proper fix and the design much more in-line with other
parts of i915.
-Daniel

> 
> - Sunil
> 
> >-----Original Message-----
> >From: Lespiau, Damien 
> >Sent: Monday, May 18, 2015 4:14 PM
> >To: Daniel Vetter
> >Cc: Konduru, Chandra; intel-gfx@lists.freedesktop.org; Shah, Suketu J; Kamath, Sunil; Manna, Animesh
> >Subject: Re: [Intel-gfx] DC6 already programmed to be disabled
> >
> >On Mon, May 18, 2015 at 10:38:03AM +0200, Daniel Vetter wrote:
> >> On Fri, May 15, 2015 at 11:22:27PM +0000, Konduru, Chandra wrote:
> >> > Hi,
> >> > I have been seeing below warning on skylake system on which dmc fw isn't placed.
> >> > Is below warning expected? If so what is it conveying?
> >> 
> >> Seems to be another fallout from the current design of how we prevent
> >> dc5/6 when the firmware is not (yet) loaded. I've detailed how this 
> >> should be fixed. We need to prevent the rpm code from ever trying to 
> >> shut down that specific power well instead of just not obeying the 
> >> request. Not obeying the request means the rpm code is out of sync 
> >> with reality, leading to WARN_ON fun like the one you've hit here.
> >
> >Hey all,
> >
> >Would anyone of you (Sunil, Animesh, Suketu) have time to fix this? (the warning when DMC firmware isn't there). We should be able to work when failing to load the DMC firmware.
> >
> >What Daniel says is not quite accurate, bear in mind we still can shut down all power wells and do PC10 with screens off when the DMC isn't loaded. We could also decide to disable run-time PM entirely when the >DMC firmware isn't there. That's something that can be fixed later on though, right now the most immediate issue is not to dump lots of warnings when failing to load the firmware.
> >
> >By default I'll fix it, I have this on my TODO list, it's quite low though.
> >
> >Thanks,
> >
> >--
> >Damien

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: DC6 already programmed to be disabled

[Sagar Arun Kamble]

The warning faced by Chandra has to corrected. This warning is not
related to firmware load. 

Issue is -> During boot DC5/6 will be already disabled and hence While
enabling PW2 for the first time the check to see if DC5/6 is already
disabled does not make sense.

So condition should be

WARN(!(I915_READ(DC_STATE_EN) & DC_STATE_EN_UPTO_DC6) && pg2_enabled),
	"DC6 already programmed to be disabled.\n");


On Tue, 2015-05-19 at 10:18 +0200, Daniel Vetter wrote:
> On Tue, May 19, 2015 at 04:53:16AM +0000, Kamath, Sunil wrote:
> > Sure Damien. We will come back with solution for the same. 
> 
> Please check out my reply to Animesh' patch in
> 
> https://tango.freedesktop.org/patch/49084/
> 
> I think this is the proper fix and the design much more in-line with other
> parts of i915.
> -Daniel
> 
> > 
> > - Sunil
> > 
> > >-----Original Message-----
> > >From: Lespiau, Damien 
> > >Sent: Monday, May 18, 2015 4:14 PM
> > >To: Daniel Vetter
> > >Cc: Konduru, Chandra; intel-gfx@lists.freedesktop.org; Shah, Suketu J; Kamath, Sunil; Manna, Animesh
> > >Subject: Re: [Intel-gfx] DC6 already programmed to be disabled
> > >
> > >On Mon, May 18, 2015 at 10:38:03AM +0200, Daniel Vetter wrote:
> > >> On Fri, May 15, 2015 at 11:22:27PM +0000, Konduru, Chandra wrote:
> > >> > Hi,
> > >> > I have been seeing below warning on skylake system on which dmc fw isn't placed.
> > >> > Is below warning expected? If so what is it conveying?
> > >> 
> > >> Seems to be another fallout from the current design of how we prevent
> > >> dc5/6 when the firmware is not (yet) loaded. I've detailed how this 
> > >> should be fixed. We need to prevent the rpm code from ever trying to 
> > >> shut down that specific power well instead of just not obeying the 
> > >> request. Not obeying the request means the rpm code is out of sync 
> > >> with reality, leading to WARN_ON fun like the one you've hit here.
> > >
> > >Hey all,
> > >
> > >Would anyone of you (Sunil, Animesh, Suketu) have time to fix this? (the warning when DMC firmware isn't there). We should be able to work when failing to load the DMC firmware.
> > >
> > >What Daniel says is not quite accurate, bear in mind we still can shut down all power wells and do PC10 with screens off when the DMC isn't loaded. We could also decide to disable run-time PM entirely when the >DMC firmware isn't there. That's something that can be fixed later on though, right now the most immediate issue is not to dump lots of warnings when failing to load the firmware.
> > >
> > >By default I'll fix it, I have this on my TODO list, it's quite low though.
> > >
> > >Thanks,
> > >
> > >--
> > >Damien
> 


_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: DC6 already programmed to be disabled

[Damien Lespiau]

On Thu, Jun 04, 2015 at 01:39:50PM +0530, Sagar Arun Kamble wrote:
> The warning faced by Chandra has to corrected. This warning is not
> related to firmware load. 
> 
> Issue is -> During boot DC5/6 will be already disabled and hence While
> enabling PW2 for the first time the check to see if DC5/6 is already
> disabled does not make sense.
> 
> So condition should be
> 
> WARN(!(I915_READ(DC_STATE_EN) & DC_STATE_EN_UPTO_DC6) && pg2_enabled),
> 	"DC6 already programmed to be disabled.\n");

Hi,

we may talk about different problems here. The warn in question here is
when the DMC firmware isn't loaded (because, for instance, it's not on
the filesystem)

-- 
Damien
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: DC6 already programmed to be disabled

[Daniel Vetter]

On Thu, Jun 04, 2015 at 10:41:42AM +0100, Damien Lespiau wrote:
> On Thu, Jun 04, 2015 at 01:39:50PM +0530, Sagar Arun Kamble wrote:
> > The warning faced by Chandra has to corrected. This warning is not
> > related to firmware load. 
> > 
> > Issue is -> During boot DC5/6 will be already disabled and hence While
> > enabling PW2 for the first time the check to see if DC5/6 is already
> > disabled does not make sense.
> > 
> > So condition should be
> > 
> > WARN(!(I915_READ(DC_STATE_EN) & DC_STATE_EN_UPTO_DC6) && pg2_enabled),
> > 	"DC6 already programmed to be disabled.\n");
> 
> Hi,
> 
> we may talk about different problems here. The warn in question here is
> when the DMC firmware isn't loaded (because, for instance, it's not on
> the filesystem)

Also the initial power domains tricks should result in us not trying to
double-disable stuff on driver load. It works under the assumption that
the bios hasn't enabled any power conservation tricks. If that assumption
ever proves wrong then we need to replace the hard-coded list of initial
power domains with a dynamically generated one. Unfortuantely that also
means fixing up our init code to properly take power domain references, so
quite a bit of work.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx