As for LPAE, it's simply a case of skipping the leaf invalidation for a regular unmap, and ensuring that the one in split_blk_unmap() is paired with an explicit sync ASAP rather than relying on one which might only eventually happen way down the line.
Signed-off-by: Robin Murphy <robin.mur...@arm.com> --- v8: Do this properly instead of just hacking around not doing it (look, it turns out barely any more complicated!) drivers/iommu/io-pgtable-arm-v7s.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/drivers/iommu/io-pgtable-arm-v7s.c b/drivers/iommu/io-pgtable-arm-v7s.c index b5948ba6b3b3..445c3bde0480 100644 --- a/drivers/iommu/io-pgtable-arm-v7s.c +++ b/drivers/iommu/io-pgtable-arm-v7s.c @@ -587,6 +587,7 @@ static size_t arm_v7s_split_blk_unmap(struct arm_v7s_io_pgtable *data, } io_pgtable_tlb_add_flush(&data->iop, iova, size, size, true); + io_pgtable_tlb_sync(&data->iop); return size; } @@ -642,6 +643,13 @@ static size_t __arm_v7s_unmap(struct arm_v7s_io_pgtable *data, io_pgtable_tlb_sync(iop); ptep = iopte_deref(pte[i], lvl); __arm_v7s_free_table(ptep, lvl + 1, data); + } else if (iop->cfg.quirks & IO_PGTABLE_QUIRK_NON_STRICT) { + /* + * Order the PTE update against queueing the IOVA, to + * guarantee that a flush callback from a different CPU + * has observed it before the TLBIALL can be issued. + */ + smp_wmb(); } else { io_pgtable_tlb_add_flush(iop, iova, blk_size, blk_size, true); @@ -712,7 +720,8 @@ static struct io_pgtable *arm_v7s_alloc_pgtable(struct io_pgtable_cfg *cfg, IO_PGTABLE_QUIRK_NO_PERMS | IO_PGTABLE_QUIRK_TLBI_ON_MAP | IO_PGTABLE_QUIRK_ARM_MTK_4GB | - IO_PGTABLE_QUIRK_NO_DMA)) + IO_PGTABLE_QUIRK_NO_DMA | + IO_PGTABLE_QUIRK_NON_STRICT)) return NULL; /* If ARM_MTK_4GB is enabled, the NO_PERMS is also expected. */ -- 2.19.0.dirty