[PPC64] Fix NUMA on POWER8
ClosedPublic
Actions

Authored by luporl on Jan 29 2020, 4:16 PM.

Details

Reviewers

Commits

Summary

On some POWER8 machines, 'ibm,associativity' property may have 6
cells, which would overflow the 5 cells buffer being used.
There was also an issue with the "check if node is root" part,
that have been fixed too.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

luporl created this revision.Jan 29 2020, 4:16 PM

Herald added subscribers: alfredo, bdragon. · View Herald TranscriptJan 29 2020, 4:16 PM

Harbormaster completed remote builds in B29015: Diff 67458.Jan 29 2020, 4:16 PM

jhibbits added inline comments.Jan 29 2020, 4:42 PM

sys/powerpc/powernv/platform_powernv.c
535–536 ↗	(On Diff #67458)	I would suggest you just move this up, and not bother with the OF_getproplen(), just do res = OF_getencprop(...) if (res < 0) { ... }

Address review comments

Harbormaster completed remote builds in B29018: Diff 67462.Jan 29 2020, 5:19 PM

luporl marked an inline comment as done.Jan 29 2020, 5:20 PM

luporl added inline comments.

sys/powerpc/powernv/platform_powernv.c
535–536 ↗	(On Diff #67458)	Right, it's better indeed.

luporl marked an inline comment as done.Jan 29 2020, 5:22 PM

jhibbits added inline comments.Jan 29 2020, 5:25 PM

sys/powerpc/powernv/platform_powernv.c
532–534 ↗	(On Diff #67462)	I think this KASSERT() is unnecessary. A violation of this "invariant" is really a broken device tree, not a kernel bug. Maybe "if (res < sizeof(cell_t) * (platform_associativity + 1))" up above instead? We can limp along with a busted device tree, we don't need to panic.