Page MenuHomeFreeBSD
Differential D26466

Add largepage support to the arm64 pmap.
ClosedPublic
Actions

Authored by markj on Sep 17 2020, 4:35 PM.
Tags
None
Referenced Files
Unknown Object (File)
Mon, Dec 30, 2:44 PM
Unknown Object (File)
Tue, Dec 17, 6:37 PM
Unknown Object (File)
Nov 25 2024, 5:53 PM
Unknown Object (File)
Nov 15 2024, 5:37 AM
Unknown Object (File)
Oct 7 2024, 5:12 PM
Unknown Object (File)
Sep 27 2024, 11:24 PM
Unknown Object (File)
Sep 24 2024, 7:43 AM
Unknown Object (File)
Sep 23 2024, 6:07 PM
View All 63 Files
Subscribers
andrew
emaste
imp

Details

Reviewers
manu
alc
kib
Group Reviewers
arm64
Commits
rS366089: Add largepage support to the arm64 pmap.

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

markj created this revision.Sep 17 2020, 4:35 PM
Herald added a reviewer: manu. · View Herald TranscriptSep 17 2020, 4:35 PM
Herald added subscribers: emaste, andrew. · View Herald Transcript
markj requested review of this revision.Sep 17 2020, 4:35 PM
Harbormaster completed remote builds in B33609: Diff 77142.Sep 17 2020, 4:35 PM
markj added reviewers: alc, kib, arm64.Sep 17 2020, 4:38 PM
markj updated this revision to Diff 77151.Sep 17 2020, 5:36 PM

Harmless typo (L1_BLOCK -> L2_BLOCK in an assertion).

Harbormaster completed remote builds in B33613: Diff 77151.Sep 17 2020, 5:36 PM
alc added a comment.Sep 17 2020, 6:44 PM

All of my recent amd64 comments apply here as well.

markj updated this revision to Diff 77191.Sep 18 2020, 5:20 PM
  • Adjust assertion messages.
  • Implement pmap_copy() support.
Harbormaster completed remote builds in B33629: Diff 77191.Sep 18 2020, 5:20 PM
alc added inline comments.Sep 19 2020, 3:18 AM
sys/arm64/arm64/pmap.c
2923 ↗(On Diff #77191)

pmap_l0(pmap, sva) -> l0

3479 ↗(On Diff #77191)

I'm confused. l1 can't have ATTR_DESCR_VALID set, but the pmap's resident count isn't being incremented here (like it is below in the else).

3503 ↗(On Diff #77191)

Can't the l0 entry be invalid? In which case, l1p will be NULL.

3516 ↗(On Diff #77191)

See my earlier comment.

4291 ↗(On Diff #77191)

Assert that the l1 entry has the "wired" bit set, like we do below?

markj updated this revision to Diff 77220.Sep 19 2020, 3:15 PM
markj marked 5 inline comments as done.

Address review feedback.

Harbormaster completed remote builds in B33647: Diff 77220.Sep 19 2020, 3:15 PM
markj added inline comments.Sep 19 2020, 3:16 PM
sys/arm64/arm64/pmap.c
3479 ↗(On Diff #77191)

You're right, I'm not sure why I didn't just follow what the amd64 implementation does.

alc added inline comments.Sep 19 2020, 6:41 PM
sys/arm64/arm64/pmap.c
1267 ↗(On Diff #77220)

Hmm. This is wrong for managed mappings. We should be testing whether the page is supposed to be writeable, not whether we've dirtied it.

alc added inline comments.Sep 19 2020, 7:16 PM
sys/arm64/arm64/pmap.c
2923–2924 ↗(On Diff #77220)

pmap_unuse_pt(pmap, sva, pmap_load(l0), &free);?

3479 ↗(On Diff #77191)

Suppose that you are replacing an existing large page mapping. Isn't there still a problem? Because you'll increment the resident count, but you shouldn't.

markj added inline comments.Sep 19 2020, 7:19 PM
sys/arm64/arm64/pmap.c
1267 ↗(On Diff #77220)

Right. We should be checking (tpte & ATTR_SW_DBM) != 0 I believe.

Does this bug impact correctness? If we fail here because the logically writable mapping is clean, then vm_fault_quick_hold_pages()'s fallback logic will invoke the fault handler and locate the page.

I'm not sure about stage2 translations yet.

markj updated this revision to Diff 77230.Sep 19 2020, 7:44 PM
markj marked 2 inline comments as done.

Address feedback:

  • Use pmap_unuse_pt().
  • Fix resident page accounting, second attempt.
Harbormaster completed remote builds in B33654: Diff 77230.Sep 19 2020, 7:44 PM
kib added inline comments.Sep 19 2020, 10:00 PM
sys/arm64/arm64/pmap.c
3496 ↗(On Diff #77230)

Don't you need to do some kind of dsb() after store, at least for the new mapping ?

alc added inline comments.Sep 19 2020, 10:11 PM
sys/arm64/arm64/pmap.c
3496 ↗(On Diff #77230)

Yes.

alc added inline comments.Sep 20 2020, 12:55 AM
sys/arm64/arm64/pmap.c
3478 ↗(On Diff #77230)

I don't understand why we do this unconditionally. We don't use mp except in the then clause of the below if statement.

3514 ↗(On Diff #77230)

Ditto.

3533 ↗(On Diff #77230)

See my amd64 comment.

markj updated this revision to Diff 77259.Sep 20 2020, 5:24 PM

Address feedback:

  • When mapping a large page, don't look up the PTP unless we need it.
  • Simplify pmap wired count updates.
  • Unconditionally issue dsb after updating a large page PTE.
Harbormaster completed remote builds in B33672: Diff 77259.Sep 20 2020, 5:24 PM
alc added inline comments.Sep 20 2020, 10:22 PM
sys/arm64/arm64/pmap.c
3520 ↗(On Diff #77259)

1G -> 2M

3529 ↗(On Diff #77259)

L1_SIZE -> pagesizes[psind]

3532 ↗(On Diff #77259)

L1_SIZE -> pagesizes[psind]

markj updated this revision to Diff 77275.Sep 21 2020, 12:33 AM

Address feedback:

  • Fix assertion message.
  • Correct wired page accounting.
Harbormaster completed remote builds in B33684: Diff 77275.Sep 21 2020, 12:33 AM
alc accepted this revision.Sep 22 2020, 6:50 PM
This revision is now accepted and ready to land.Sep 22 2020, 6:50 PM
kib accepted this revision.Sep 23 2020, 5:36 PM
Closed by commit rS366089: Add largepage support to the arm64 pmap. (authored by markj). · Explain WhySep 23 2020, 7:34 PM
This revision was automatically updated to reflect the committed changes.
markj added a commit: rS366089: Add largepage support to the arm64 pmap..
Herald added a subscriber: imp. · View Herald TranscriptSep 23 2020, 7:34 PM

Revision Contents
Changeset List

PathSize
head/
sys/
arm64/
arm64/
226 lines
include/
2 lines

Diff 77452

View Options

head/sys/arm64/arm64/pmap.c

Loading...
View Options

head/sys/arm64/include/param.h

Loading...