6.15-stable review patch. If anyone has any objections, please let me know.
------------------
From: Tejas Upadhyay tejas.upadhyay@intel.com
[ Upstream commit fd25fa90edcfd4db5bf69c11621021a7cfd11d53 ]
Skipping TLB invalidations on VF causing unrecoverable faults. Probable reason for skipping TLB invalidations on SRIOV could be lack of support for instruction MI_FLUSH_DW_STORE_INDEX. Add back TLB flush with some additional handling.
Helps in resolving, [ 704.913454] xe 0000:00:02.1: [drm:pf_queue_work_func [xe]] ASID: 0 VFID: 0 PDATA: 0x0d92 Faulted Address: 0x0000000002fa0000 FaultType: 0 AccessType: 1 FaultLevel: 0 EngineClass: 3 bcs EngineInstance: 8 [ 704.913551] xe 0000:00:02.1: [drm:pf_queue_work_func [xe]] Fault response: Unsuccessful -22
V2: - Use Xmas tree (MichalW)
Suggested-by: Matthew Brost matthew.brost@intel.com Fixes: 97515d0b3ed92 ("drm/xe/vf: Don't emit access to Global HWSP if VF") Reviewed-by: Matthew Brost matthew.brost@intel.com Link: https://patchwork.freedesktop.org/patch/msgid/20250710045945.1023840-1-tejas... Signed-off-by: Tejas Upadhyay tejas.upadhyay@intel.com (cherry picked from commit b528e896fa570844d654b5a4617a97fa770a1030) Signed-off-by: Lucas De Marchi lucas.demarchi@intel.com Signed-off-by: Sasha Levin sashal@kernel.org --- drivers/gpu/drm/xe/xe_ring_ops.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/xe/xe_ring_ops.c b/drivers/gpu/drm/xe/xe_ring_ops.c index bc1689db4cd71..7b50c7c1ee21d 100644 --- a/drivers/gpu/drm/xe/xe_ring_ops.c +++ b/drivers/gpu/drm/xe/xe_ring_ops.c @@ -110,13 +110,14 @@ static int emit_bb_start(u64 batch_addr, u32 ppgtt_flag, u32 *dw, int i) return i; }
-static int emit_flush_invalidate(u32 *dw, int i) +static int emit_flush_invalidate(u32 addr, u32 val, u32 *dw, int i) { dw[i++] = MI_FLUSH_DW | MI_INVALIDATE_TLB | MI_FLUSH_DW_OP_STOREDW | - MI_FLUSH_IMM_DW | MI_FLUSH_DW_STORE_INDEX; - dw[i++] = LRC_PPHWSP_FLUSH_INVAL_SCRATCH_ADDR; - dw[i++] = 0; + MI_FLUSH_IMM_DW; + + dw[i++] = addr | MI_FLUSH_DW_USE_GTT; dw[i++] = 0; + dw[i++] = val;
return i; } @@ -397,23 +398,20 @@ static void __emit_job_gen12_render_compute(struct xe_sched_job *job, static void emit_migration_job_gen12(struct xe_sched_job *job, struct xe_lrc *lrc, u32 seqno) { + u32 saddr = xe_lrc_start_seqno_ggtt_addr(lrc); u32 dw[MAX_JOB_SIZE_DW], i = 0;
i = emit_copy_timestamp(lrc, dw, i);
- i = emit_store_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc), - seqno, dw, i); + i = emit_store_imm_ggtt(saddr, seqno, dw, i);
dw[i++] = MI_ARB_ON_OFF | MI_ARB_DISABLE; /* Enabled again below */
i = emit_bb_start(job->ptrs[0].batch_addr, BIT(8), dw, i);
- if (!IS_SRIOV_VF(gt_to_xe(job->q->gt))) { - /* XXX: Do we need this? Leaving for now. */ - dw[i++] = preparser_disable(true); - i = emit_flush_invalidate(dw, i); - dw[i++] = preparser_disable(false); - } + dw[i++] = preparser_disable(true); + i = emit_flush_invalidate(saddr, seqno, dw, i); + dw[i++] = preparser_disable(false);
i = emit_bb_start(job->ptrs[1].batch_addr, BIT(8), dw, i);