Bug 67734

Your description is very, very confused. Please do copy and paste the exact sequence of commands you use, and then describe exactly what you expected.

Always attach a dmesg.

Jesse is the Batrail owner.

Chris, this bug is reported for  GFX turbo feature on JIRA, owner is Jesse who provided us test methodology, he must know about it. 
(In reply to comment #1)
> Your description is very, very confused. Please do copy and paste the exact
> sequence of commands you use, and then describe exactly what you expected.
> 
> Always attach a dmesg.

The bug report is still imprecise.

(In reply to comment #4)
> The bug report is still imprecise.

Let me clarify the reproduced steps,
1. Run a workload(Like as Openarena v0.8.8)
2. Navigate to /sys/class/drm/card0 to check P-state frequency
3. Observe some files indicated P-state frequency(gt_cur_freq_mhz is the current frequency, min/max are the min/max freqs, RP0 is the max P-state, RP1 is the "nominal" P-state, and RPn is the lowest frequency)

Actual: The value of RP0/RP1/RPn frequency returned from these files is equal to 0.
Expected: It should be max/nominal/max freq.

(In reply to comment #5)
> (In reply to comment #4)
> > The bug report is still imprecise.
> 
> Let me clarify the reproduced steps,
> 1. Run a workload(Like as Openarena v0.8.8)
> 2. Navigate to /sys/class/drm/card0 to check P-state frequency
> 3. Observe some files indicated P-state frequency(gt_cur_freq_mhz is the
> current frequency, min/max are the min/max freqs, RP0 is the max P-state,
> RP1 is the "nominal" P-state, and RPn is the lowest frequency)
> 
> Actual: The value of RP0/RP1/RPn frequency returned from these files is
> equal to 0.
> Expected: It should be max/nominal/max freq.

The purpose is to check whether P-state frequency will increase or not while running a workload.

Hm, vlv rps support was merged for 3.11 ... I have no idea why this doesn't work. Jesse?

rps support is there. Just gt_rp_mhz_show() was not adjusted for VLV.

Created attachment 83707 [details] [review]
Expose RPe for vlv in sysfs

(In reply to comment #9)
> Created attachment 83707 [details] [review] [review]
> Expose RPe for vlv in sysfs

Chris, I can't find this posted on the ml. But you can stick my r-b on it when you do.

Was waiting for the bug owner to notice ;-)

Assigning to Kobe. Please confirm.

(In reply to comment #9)
> Created attachment 83707 [details] [review] [review]
> Expose RPe for vlv in sysfs

Chris, I can't find gt_RP0_freq_mhz/gt_RP1_freq_mhz/gt_RPn_freq_mhz under /sys/class/drm/card0

[root@x-byt02 card0]# ls
card0-DP-1      card0-HDMI-A-2  dev     error            gt_max_freq_mhz  power      uevent
card0-HDMI-A-1  card0-VGA-1     device  gt_cur_freq_mhz  gt_min_freq_mhz  subsystem  vlv_rpe_freq_mhz

Baytrail doesn't use RP0, RP1, or RPn. Hence the patch does not expose the sysfs files.

(In reply to comment #14)
> Baytrail doesn't use RP0, RP1, or RPn. Hence the patch does not expose the
> sysfs files.


All right, if so, this bug can be closed.

The patch isn't merged yet, so we need to keep the bug report open until the fix has landed.

commit 97e4eed7dc532e20208b0bdf7ad1136569da2f35
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Aug 26 16:18:54 2013 +0100

    drm/i915: Adjust available RPS information through sysfs for vlv
    
    Valleyview has its own render power state implementation with different
    capability knobs - it has no RP0,RP1,RPn but rather RPe.
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67734
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
    Tested-by: kobe.qin@intel.com
    Reviewed-by: Jani Nikula <jani.nikula@intel.com>
    Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Closing old verified+fixed.