Try to handle TLB invalidations from CT fast-path
Get series with b4
b4 am [email protected]Versions newest first · linked by author and title
- v6 Latest11 patches 2 repliesIn review
- v510 patches 4 repliesSuperseded
- v47 patches 15 repliesSuperseded
Earlier versions · 3
| Version | Posted | Patches | Replies | Status |
|---|---|---|---|---|
| v3 | 5 patches | 2 replies | Superseded | |
| v2 | 1 patch | 4 replies | Superseded | |
| v1 | 1 patch | 1 reply | Superseded |
Changes in v6 from the cover letter
- One small fix to patch 8 to make sure we write the correct seqno on reset.
Patches in v6 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/11 | drm/xe/tlb: drop unnecessary smp_wmb()drivers/ | +4 | 0 replies | |
| 2/11 | drm/xe/tlb: ensure we access seqno_recv oncedrivers/ | +6 | 0 replies | |
| 3/11 | drm/xe: hold mem_access.ref for CT fast-pathdrivers/ drivers/ drivers/ | +10 | 0 replies | |
| 4/11 | drm/xe/ct: hold fast_lock when reserving space for g2hdrivers/ | +15 | 0 replies | |
| 5/11 | drm/xe/tlb: increment next seqno after successful CT senddrivers/ | +6 | 0 replies | |
| 6/11 | drm/xe/ct: serialise fast_lock during CT disabledrivers/ | +6 | 0 replies | |
| 7/11 | drm/xe/gt: tweak placement for signalling TLB fences after GT resetdrivers/ | +2 | 0 replies | |
| 8/11 | drm/xe/tlb: also update seqno_recv during resetdrivers/ | +22 | 0 replies | |
| 9/11 | drm/xe/tlb: print seqno_recv on fence TLB timeoutdrivers/ | +2 | 0 replies | |
| 10/11 | drm/xe/ct: update g2h outstanding for CTB capturedrivers/ | +1 | 1 reply | |
| 11/11 | drm/xe: handle TLB invalidations from CT fast-pathdrivers/ drivers/ drivers/ | +91 | 0 replies |
Reviews trailers per patch
All 11 patches have a Reviewed-by or Acked-by.
| Patch | Matthew Brost | José Roberto de Souza |
|---|---|---|
| Series | T-b | |
| 1/11 | R-b | |
| 2/11 | R-b | |
| 3/11 | R-b | |
| 4/11 | R-b | |
| 5/11 | R-b | |
| 6/11 | R-b | |
| 7/11 | R-b | |
| 8/11 | R-b | |
| 9/11 | R-b | |
| 10/11 | R-b | |
| 11/11 | R-b |
Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.
Cover letter [Intel-xe] [PATCH v6 00/11] Try to handle TLB invalidations from CT fast-path
In various test cases and workloads that put the system under a heavy load, we can sometimes see errors with missed TLB invalidations. If we handle the TLB completions directly from IRQ we can see a big improvement.
v4:
- Various tweaks. Big thing is to make sure we grab the fast_lock for the g2h reservation when doing the CT send, which closes one big race with the CT fast-path.
v5:
- Various improvements. Fix the race with adding a fence after its seqno has already been signalled. (Matt Brost)
v6:
- One small fix to patch 8 to make sure we write the correct seqno on reset.
Signature · 1 line
2.41.0