From: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx> [ Upstream commit 28a77185f1cd0650b664f54614143aaaa3a7a615 ] It is incorrect to always clear PRO when it's set w/o first checking whether the overflow condition has been cleared. Current code assumes that if an overflow condition occurs it must have been cleared by earlier loop. However since the code runs in a threaded context, the overflow condition could occur even after setting the head to the tail under some extreme condition. To be sane, we should read both head/tail again when seeing a pending PRO and only clear PRO after all pending PRs have been handled. Suggested-by: Kevin Tian <kevin.tian@xxxxxxxxx> Signed-off-by: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx> Link: https://lore.kernel.org/linux-iommu/MWHPR11MB18862D2EA5BD432BF22D99A48CA09@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/ Link: https://lore.kernel.org/r/20210126080730.2232859-2-baolu.lu@xxxxxxxxxxxxxxx Signed-off-by: Joerg Roedel <jroedel@xxxxxxx> Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx> --- drivers/iommu/intel/svm.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) diff --git a/drivers/iommu/intel/svm.c b/drivers/iommu/intel/svm.c index 43f392d27d31..b200a3acc6ed 100644 --- a/drivers/iommu/intel/svm.c +++ b/drivers/iommu/intel/svm.c @@ -1079,8 +1079,17 @@ prq_advance: * Clear the page request overflow bit and wake up all threads that * are waiting for the completion of this handling. */ - if (readl(iommu->reg + DMAR_PRS_REG) & DMA_PRS_PRO) - writel(DMA_PRS_PRO, iommu->reg + DMAR_PRS_REG); + if (readl(iommu->reg + DMAR_PRS_REG) & DMA_PRS_PRO) { + pr_info_ratelimited("IOMMU: %s: PRQ overflow detected\n", + iommu->name); + head = dmar_readq(iommu->reg + DMAR_PQH_REG) & PRQ_RING_MASK; + tail = dmar_readq(iommu->reg + DMAR_PQT_REG) & PRQ_RING_MASK; + if (head == tail) { + writel(DMA_PRS_PRO, iommu->reg + DMAR_PRS_REG); + pr_info_ratelimited("IOMMU: %s: PRQ overflow cleared", + iommu->name); + } + } if (!completion_done(&iommu->prq_complete)) complete(&iommu->prq_complete); -- 2.30.1
