From: "Aneesh Kumar K.V" <aneesh.kumar@xxxxxxxxxxxxx> [ Upstream commit 8f53f9c0f68ab2168f637494b9e24034899c1310 ] With a 64K page size flush with start and end: (start, end) = (721f680d0000, 721f680e0000) results in: (hstart, hend) = (721f68200000, 721f68000000) ie. hstart is above hend, which indicates no huge page flush is needed. However the current logic incorrectly sets hflush = true in this case, because hstart != hend. That causes us to call __tlbie_va_range() passing hstart/hend, to do a huge page flush even though we don't need to. __tlbie_va_range() will skip the actual tlbie operation for start > end. But it will still end up calling fixup_tlbie_va_range() and doing the TLB fixups in there, which is harmless but unnecessary work. Reported-by: Bharata B Rao <bharata@xxxxxxxxxxxxx> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@xxxxxxxxxxxxx> Reviewed-by: Nicholas Piggin <npiggin@xxxxxxxxx> [mpe: Drop else case, hflush is already false, flesh out change log] Signed-off-by: Michael Ellerman <mpe@xxxxxxxxxxxxxx> Link: https://lore.kernel.org/r/20200513030616.152288-1-aneesh.kumar@xxxxxxxxxxxxx Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx> --- arch/powerpc/mm/book3s64/radix_tlb.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/arch/powerpc/mm/book3s64/radix_tlb.c b/arch/powerpc/mm/book3s64/radix_tlb.c index 758ade2c2b6e..b5cc9b23cf02 100644 --- a/arch/powerpc/mm/book3s64/radix_tlb.c +++ b/arch/powerpc/mm/book3s64/radix_tlb.c @@ -884,9 +884,7 @@ static inline void __radix__flush_tlb_range(struct mm_struct *mm, if (IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE)) { hstart = (start + PMD_SIZE - 1) & PMD_MASK; hend = end & PMD_MASK; - if (hstart == hend) - hflush = false; - else + if (hstart < hend) hflush = true; } -- 2.25.1
