AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Jay Cornwall <jay.cornwall@amd.com>
To: <amd-gfx@lists.freedesktop.org>
Cc: Lancelot Six <lancelot.six@amd.com>,
	Alexey Kondratiev <Alexey.Kondratiev@amd.com>,
	Jay Cornwall <jay.cornwall@amd.com>,
	"Vladimir Indic" <vladimir.indic@amd.com>
Subject: [PATCH 5/5] drm/amdkfd: Do not include VGPR MSBs in saved PC during save
Date: Fri, 16 Jan 2026 14:39:32 -0600	[thread overview]
Message-ID: <20260116203932.988704-6-jay.cornwall@amd.com> (raw)
In-Reply-To: <20260116203932.988704-1-jay.cornwall@amd.com>

From: Lancelot Six <lancelot.six@amd.com>

The current trap handler uses the top bits of ttmp1 to store a copy of
sq_wave_mode.*vgpr_msb (except for src2_vgpr_msb).  This is so the
effective values in sq_wave_mode can be cleared to ensure correct
behavior of the trap handler.

When saving sq_wave_mode, the trap handler correctly rebuilds the
expected value (with *vgpr_msb restored), so the save area is correct.
However, the PC itself is copied from ttmp[0:1], which contains the
wave's PC as well as the saved MSBs.

The debugger reads the PC from the save area and is confused when non-0
values from VGPR_MSBs are present.

This patch fixes this by saving the PC in the save area's PC slot, not
the composite of the PC and VGPR_MSBs.  On restore, the VGPR_MSBs are
restored from sq_wave_mode.

Signed-off-by: Lancelot Six <lancelot.six@amd.com>
Tested-by: Alexey Kondratiev <Alexey.Kondratiev@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Cc: Vladimir Indic <vladimir.indic@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h         | 6 +++---
 drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
index 9bb7fb6a83ed..39bdc98b8b6d 100644
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
@@ -3760,8 +3760,8 @@ static const uint32_t cwsr_trap_gfx12_hex[] = {
 	0xb8faf804, 0x8b7a847a,
 	0x91788478, 0x8c787a78,
 	0xd7610002, 0x0000fa6c,
-	0x807d817d, 0x917aff6d,
-	0x80000000, 0xd7610002,
+	0x807d817d, 0x8b7aff6d,
+	0x0000ffff, 0xd7610002,
 	0x0000fa7a, 0x807d817d,
 	0xd7610002, 0x0000fa6e,
 	0x807d817d, 0xd7610002,
@@ -4848,7 +4848,7 @@ static const uint32_t cwsr_trap_gfx12_1_0_hex[] = {
 	0x9178ff78, 0x0001000c,
 	0x8c787a78, 0xd7610002,
 	0x0000fa6c, 0x807d817d,
-	0x917aff6d, 0x80000000,
+	0x8b7aff6d, 0x01ffffff,
 	0xd7610002, 0x0000fa7a,
 	0x807d817d, 0xd7610002,
 	0x0000fa6e, 0x807d817d,
diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm
index ccc61f60ceb3..c33e7660d8f4 100644
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm
@@ -544,7 +544,7 @@ L_SAVE_HWREG:
 	s_or_b32	s_save_state_priv, s_save_state_priv, s_save_tmp
 
 	write_hwreg_to_v2(s_save_pc_lo)
-	s_andn2_b32	s_save_tmp, s_save_pc_hi, S_SAVE_PC_HI_FIRST_WAVE_MASK
+	s_and_b32       s_save_tmp, s_save_pc_hi, ADDRESS_HI32_MASK
 	write_hwreg_to_v2(s_save_tmp)
 	write_hwreg_to_v2(s_save_exec_lo)
 #if WAVE32_ONLY
-- 
2.34.1


      parent reply	other threads:[~2026-01-16 20:40 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-01-16 20:39 [PATCH 0/5] drm/amdkfd: Trap handler fixes and gfx12.1 support Jay Cornwall
2026-01-16 20:39 ` [PATCH 1/5] drm/amdkfd: Sync trap handler binary with source Jay Cornwall
2026-01-20 22:34   ` Lancelot SIX
2026-01-21 10:27     ` Indic, Vladimir
2026-01-16 20:39 ` [PATCH 2/5] drm/amdkfd: Fix scalar load ordering in gfx12.1 trap handler Jay Cornwall
2026-01-20 22:38   ` Lancelot SIX
2026-01-21 10:32     ` Indic, Vladimir
2026-01-16 20:39 ` [PATCH 3/5] drm/amdkfd: gfx12.1 cluster barrier context save workaround Jay Cornwall
2026-01-20 23:27   ` Lancelot SIX
2026-01-21 10:37     ` Indic, Vladimir
2026-01-16 20:39 ` [PATCH 4/5] drm/amdkfd: gfx12.1 trap handler support for expert scheduling mode Jay Cornwall
2026-01-20 23:30   ` Lancelot SIX
2026-01-21 10:46     ` Indic, Vladimir
2026-01-16 20:39 ` Jay Cornwall [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260116203932.988704-6-jay.cornwall@amd.com \
    --to=jay.cornwall@amd.com \
    --cc=Alexey.Kondratiev@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=lancelot.six@amd.com \
    --cc=vladimir.indic@amd.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox