On 25/08/2022 18:49, Sripada, Radhakrishna wrote:
Hi G.G,
-----Original Message-----
From: Mun, Gwan-gyeong <gwan-gyeong.mun@xxxxxxxxx>
Sent: Tuesday, August 23, 2022 11:14 PM
To: Roper, Matthew D <matthew.d.roper@xxxxxxxxx>; Sripada, Radhakrishna
<radhakrishna.sripada@xxxxxxxxx>
Cc: intel-gfx@xxxxxxxxxxxxxxxxxxxxx
Subject: Re: [PATCH] drm/i915: Skip Bit12 fw domain reset for gen12+
On 8/18/22 3:00 PM, Matt Roper wrote:
On Wed, Aug 17, 2022 at 03:43:04PM -0700, Radhakrishna Sripada wrote:
Bit12 of the Forcewake request register should not be cleared post
gen12. Do not touch this bit while clearing during fw domain reset.
Bspec: 52542
Signed-off-by: Sushma Venkatesh Reddy
<sushma.venkatesh.reddy@xxxxxxxxx>
Signed-off-by: Radhakrishna Sripada <radhakrishna.sripada@xxxxxxxxx>
---
drivers/gpu/drm/i915/intel_uncore.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/intel_uncore.c
b/drivers/gpu/drm/i915/intel_uncore.c
index a852c471d1b3..c85e2b686c95 100644
--- a/drivers/gpu/drm/i915/intel_uncore.c
+++ b/drivers/gpu/drm/i915/intel_uncore.c
@@ -113,7 +113,10 @@ fw_domain_reset(const struct
intel_uncore_forcewake_domain *d)
* off in ICL+), so no waiting for acks
*/
/* WaRsClearFWBitsAtReset:bdw,skl */
While we're at it, let's remove the "bdw,skl" from this comment since
it's misleading and doesn't match the code. We do still apply this
workaround on other pre-gen12 platforms than just those two.
Aside from the comment tweak,
Reviewed-by: Matt Roper <matthew.d.roper@xxxxxxxxx>
- fw_clear(d, 0xffff);
+ if (GRAPHICS_VER(d->uncore->i915) >= 12)
Hi Radhakrishna Sripada,
In bspec 52542, there is an explanation that BIT12 should not be set for
address 0xA188 corresponding to FORCEWAKE_MT/FORCEWAKE_GT_GEN9, but
in
bspec 52466, there is no explanation that BIT12 should not be set for
address 0xA278, address of FORCEWAKE_RENDER_GEN9.
I ask if fw_domain_reset() should perform fw_clear() by comparing not
only GRAPHICS_VER() >= 12 but also checking of FW_DOMAIN_ID_RENDER and
FW_DOMAIN_ID_GT values.
Based on the note in 52542, all other WA notes are overridden by the comment. So unless stated
otherwise, it should apply to this register as well.
Created a bspec issue to request for additional clarification just to be safe. Will send an additional
patch if the comment contradicts our understanding.
How important was this patch - should it be sent to stable?
Regards,
Tvrtko
