[Bug 108336] [CI][DRMTIP] igt@*crc tests* - warn / dmesg-warn - Suspicious CRC: it looks like the CRC read back was from a register in a powered down well

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=108336

--- Comment #8 from Stanislav Lisovskiy <stanislav.lisovskiy at intel.com> ---
Found one more interesting thing: 

Sometimes wm levels are written as 0 for all levels, for planes which are
currently enabled and have non-zero ddb allocation, which at least looks
controversial. This is printed from skl_write_wm_level function, which does the
actual write to corresponding watermark level register:

 2445.635172] [PLANE:30:plane 1A] ddb (0 - 992) -> (0 - 248)
[ 2445.666098] [PLANE:58:plane 5A] ddb (0 - 0) -> (248 - 496)
[ 2445.682616] [PLANE:65:plane 6A] ddb (0 - 0) -> (496 - 744)
[ 2445.699144] [PLANE:72:plane 7A] ddb (0 - 0) -> (744 - 992)
[ 2445.718131] Writing plane 0 level 5 0 blocks while plane or ddb(f8) enabled
[ 2445.718133] Writing plane 0 level 6 0 blocks while plane or ddb(f8) enabled
[ 2445.739075] Writing plane 0 level 7 0 blocks while plane or ddb(f8) enabled
[ 2445.760015] Writing plane 4 level 5 0 blocks while plane or ddb(1f0) enabled
[ 2445.780954] Writing plane 4 level 6 0 blocks while plane or ddb(1f0) enabled
[ 2445.802147] Writing plane 4 level 7 0 blocks while plane or ddb(1f0) enabled
[ 2445.823343] Writing plane 5 level 5 0 blocks while plane or ddb(2e8) enabled
[ 2445.844534] Writing plane 5 level 6 0 blocks while plane or ddb(2e8) enabled
[ 2445.865725] Writing plane 5 level 7 0 blocks while plane or ddb(2e8) enabled
[ 2445.886923] Writing plane 6 level 5 0 blocks while plane or ddb(3e0) enabled
[ 2445.908116] Writing plane 6 level 6 0 blocks while plane or ddb(3e0) enabled
[ 2445.929306] Writing plane 6 level 7 0 blocks while plane or ddb(3e0) enabled

So we are writing 0 blocks for enabled plane, which has non-zero ddb(ddb_y->end
in brackets) allocation. I wonder if this is correct at all as ddb allocation
currently seems to be calculated based on watermarks, moreover does it even
make sense to have 0 wm level blocks and non zero ddb range for that?

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20190114/6fdea9e0/attachment.html>