From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joerg Roedel Subject: [PATCH 5/5] iommu/amd: Handle parallel invalidate_range_start/end calls correctly Date: Tue, 20 May 2014 23:18:26 +0200 Message-ID: <1400620706-21031-6-git-send-email-joro@8bytes.org> References: <1400620706-21031-1-git-send-email-joro@8bytes.org> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1400620706-21031-1-git-send-email-joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org Errors-To: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org To: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org Cc: Jay.Cornwall-5C7GfCeVMHo@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Joerg Roedel List-Id: iommu@lists.linux-foundation.org From: Joerg Roedel Add a counter to the pasid_state so that we do not restore the original page-table before all invalidate_range_start to invalidate_range_end sections have finished. Signed-off-by: Joerg Roedel --- drivers/iommu/amd_iommu_v2.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) diff --git a/drivers/iommu/amd_iommu_v2.c b/drivers/iommu/amd_iommu_v2.c index d9309a0..9ed9da2 100644 --- a/drivers/iommu/amd_iommu_v2.c +++ b/drivers/iommu/amd_iommu_v2.c @@ -45,6 +45,8 @@ struct pri_queue { struct pasid_state { struct list_head list; /* For global state-list */ atomic_t count; /* Reference count */ + atomic_t mmu_notifier_count; /* Counting nested mmu_notifier + calls */ struct task_struct *task; /* Task bound to this PASID */ struct mm_struct *mm; /* mm_struct for the faults */ struct mmu_notifier mn; /* mmu_otifier handle */ @@ -433,8 +435,11 @@ static void mn_invalidate_range_start(struct mmu_notifier *mn, pasid_state = mn_to_state(mn); dev_state = pasid_state->device_state; - amd_iommu_domain_set_gcr3(dev_state->domain, pasid_state->pasid, - __pa(empty_page_table)); + if (atomic_add_return(1, &pasid_state->mmu_notifier_count) == 1) { + amd_iommu_domain_set_gcr3(dev_state->domain, + pasid_state->pasid, + __pa(empty_page_table)); + } } static void mn_invalidate_range_end(struct mmu_notifier *mn, @@ -447,8 +452,11 @@ static void mn_invalidate_range_end(struct mmu_notifier *mn, pasid_state = mn_to_state(mn); dev_state = pasid_state->device_state; - amd_iommu_domain_set_gcr3(dev_state->domain, pasid_state->pasid, - __pa(pasid_state->mm->pgd)); + if (atomic_dec_and_test(&pasid_state->mmu_notifier_count)) { + amd_iommu_domain_set_gcr3(dev_state->domain, + pasid_state->pasid, + __pa(pasid_state->mm->pgd)); + } } static void mn_release(struct mmu_notifier *mn, struct mm_struct *mm) @@ -640,6 +648,7 @@ int amd_iommu_bind_pasid(struct pci_dev *pdev, int pasid, goto out; atomic_set(&pasid_state->count, 1); + atomic_set(&pasid_state->mmu_notifier_count, 0); init_waitqueue_head(&pasid_state->wq); spin_lock_init(&pasid_state->lock); -- 1.7.9.5