public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Lu Baolu <baolu.lu@linux.intel.com>
To: Jason Gunthorpe <jgg@ziepe.ca>, Kevin Tian <kevin.tian@intel.com>,
	Joerg Roedel <joro@8bytes.org>, Will Deacon <will@kernel.org>,
	Robin Murphy <robin.murphy@arm.com>,
	Jean-Philippe Brucker <jean-philippe@linaro.org>,
	Nicolin Chen <nicolinc@nvidia.com>, Yi Liu <yi.l.liu@intel.com>,
	Jacob Pan <jacob.jun.pan@linux.intel.com>
Cc: iommu@lists.linux.dev, linux-kselftest@vger.kernel.org,
	virtualization@lists.linux-foundation.org,
	linux-kernel@vger.kernel.org, Lu Baolu <baolu.lu@linux.intel.com>
Subject: [RFC PATCHES 13/17] iommufd: Add a timer for each iommufd fault data
Date: Tue, 30 May 2023 13:37:20 +0800	[thread overview]
Message-ID: <20230530053724.232765-14-baolu.lu@linux.intel.com> (raw)
In-Reply-To: <20230530053724.232765-1-baolu.lu@linux.intel.com>

In case that user space failed to read or respond the pending faults. As
the per-fault iommufd data will be possibly accessed in two different
contexts: user reading/responding and the timer expiring, add a reference
counter for each iommufd fault data and free the data only after all the
reference counters are released.

The page fault response timeout value is device-specific and indicates how
long the bus/device will wait for a response to a page fault request. The
timeout value is added to the per-device fault cookie. Ideally, it should
be calculated according to the platform configuration (PCI, ACPI, device
tree, etc.). This defines a default value of 1 second in case that no
platform opt-in is available. This default value is roughly estimated and
subject to be changed according to real use cases.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
 drivers/iommu/iommufd/iommufd_private.h |  8 +++
 drivers/iommu/iommufd/device.c          |  3 +
 drivers/iommu/iommufd/hw_pagetable.c    | 80 +++++++++++++++++++++++--
 3 files changed, 87 insertions(+), 4 deletions(-)

diff --git a/drivers/iommu/iommufd/iommufd_private.h b/drivers/iommu/iommufd/iommufd_private.h
index 0985e83a611f..f5b8a53044c4 100644
--- a/drivers/iommu/iommufd/iommufd_private.h
+++ b/drivers/iommu/iommufd/iommufd_private.h
@@ -249,9 +249,12 @@ struct hw_pgtable_fault {
 struct iommufd_fault {
 	struct device *dev;
 	ioasid_t pasid;
+	struct iommufd_hw_pagetable *hwpt;
 	struct iommu_hwpt_pgfault fault;
 	/* List head at hw_pgtable_fault:deliver or response */
 	struct list_head item;
+	struct timer_list timer;
+	refcount_t users;
 };
 
 /*
@@ -336,6 +339,11 @@ struct iommufd_device {
 
 struct iommufd_fault_cookie {
 	struct iommufd_device *idev;
+	/*
+	 * The maximum number of milliseconds that a device will wait for a
+	 * response to a page fault request.
+	 */
+	unsigned long timeout;
 };
 
 static inline struct iommufd_device *
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c
index 3408f1fc3e9f..6ad46638f4e1 100644
--- a/drivers/iommu/iommufd/device.c
+++ b/drivers/iommu/iommufd/device.c
@@ -374,6 +374,8 @@ static int iommufd_group_setup_msi(struct iommufd_group *igroup,
 	return 0;
 }
 
+#define IOMMUFD_DEFAULT_IOPF_TIMEOUT	1000
+
 static int iommufd_device_set_fault_cookie(struct iommufd_hw_pagetable *hwpt,
 					   struct iommufd_device *idev,
 					   ioasid_t pasid)
@@ -387,6 +389,7 @@ static int iommufd_device_set_fault_cookie(struct iommufd_hw_pagetable *hwpt,
 	if (!fcookie)
 		return -ENOMEM;
 	fcookie->idev = idev;
+	fcookie->timeout = IOMMUFD_DEFAULT_IOPF_TIMEOUT;
 
 	curr = iommu_set_device_fault_cookie(idev->dev, pasid, fcookie);
 	if (IS_ERR(curr)) {
diff --git a/drivers/iommu/iommufd/hw_pagetable.c b/drivers/iommu/iommufd/hw_pagetable.c
index c1f3ebdce796..8c441fd72e1f 100644
--- a/drivers/iommu/iommufd/hw_pagetable.c
+++ b/drivers/iommu/iommufd/hw_pagetable.c
@@ -6,6 +6,7 @@
 #include <linux/eventfd.h>
 #include <linux/file.h>
 #include <linux/anon_inodes.h>
+#include <linux/timer.h>
 #include <uapi/linux/iommufd.h>
 
 #include "../iommu-priv.h"
@@ -396,6 +397,60 @@ static void iommufd_compose_fault_message(struct iommu_fault *fault,
 	hwpt_fault->private_data[1] = fault->prm.private_data[1];
 }
 
+static void drain_iopf_fault(struct iommufd_fault *ifault)
+{
+	struct iommu_page_response resp = {
+		.version	= IOMMU_PAGE_RESP_VERSION_1,
+		.pasid		= ifault->fault.pasid,
+		.grpid		= ifault->fault.grpid,
+		.code		= IOMMU_PAGE_RESP_FAILURE,
+	};
+
+	if (!(ifault->fault.flags & IOMMU_FAULT_PAGE_REQUEST_LAST_PAGE))
+		return;
+
+	if ((ifault->fault.flags & IOMMU_FAULT_PAGE_REQUEST_PASID_VALID) &&
+	    (ifault->fault.flags & IOMMU_FAULT_PAGE_RESPONSE_NEEDS_PASID))
+		resp.flags = IOMMU_PAGE_RESP_PASID_VALID;
+
+	iommu_page_response(ifault->dev, &resp);
+}
+
+static void iommufd_put_fault(struct iommufd_fault *ifault)
+{
+	if (!ifault)
+		return;
+
+	if (refcount_dec_and_test(&ifault->users))
+		kfree(ifault);
+}
+
+static int iommufd_fault_timer_teardown(struct iommufd_fault *ifault)
+{
+	int rc;
+
+	rc = timer_delete(&ifault->timer);
+	if (rc)
+		iommufd_put_fault(ifault);
+
+	return rc;
+}
+
+static void iopf_timer_func(struct timer_list *t)
+{
+	struct iommufd_fault *ifault = from_timer(ifault, t, timer);
+	struct hw_pgtable_fault *fault = ifault->hwpt->fault;
+
+	mutex_lock(&fault->mutex);
+	if (!list_empty(&ifault->item)) {
+		list_del_init(&ifault->item);
+		drain_iopf_fault(ifault);
+	}
+	mutex_unlock(&fault->mutex);
+
+	iommufd_put_fault(ifault);
+}
+
 static enum iommu_page_response_code
 iommufd_hw_pagetable_iopf_handler(struct iommu_fault *fault,
 				  struct device *dev, void *data)
@@ -416,6 +471,10 @@ iommufd_hw_pagetable_iopf_handler(struct iommu_fault *fault,
 	iommufd_compose_fault_message(fault, &ifault->fault, cookie->idev->obj.id);
 	ifault->dev = dev;
 	ifault->pasid = fault->prm.pasid;
+	ifault->hwpt = hwpt;
+	refcount_set(&ifault->users, 2);
+	timer_setup(&ifault->timer, iopf_timer_func, 0);
+	mod_timer(&ifault->timer, jiffies + msecs_to_jiffies(cookie->timeout));
 
 	mutex_lock(&hwpt->fault->mutex);
 	list_add_tail(&ifault->item, &hwpt->fault->deliver);
@@ -443,10 +502,12 @@ static ssize_t hwpt_fault_fops_read(struct file *filep, char __user *buf,
 			break;
 		done += fault_size;
 		list_del_init(&ifault->item);
-		if (ifault->fault.flags & IOMMU_FAULT_PAGE_REQUEST_LAST_PAGE)
+		if (ifault->fault.flags & IOMMU_FAULT_PAGE_REQUEST_LAST_PAGE) {
 			list_add_tail(&ifault->item, &fault->response);
-		else
-			kfree(ifault);
+		} else {
+			iommufd_fault_timer_teardown(ifault);
+			iommufd_put_fault(ifault);
+		}
 	}
 	mutex_unlock(&fault->mutex);
 
@@ -526,6 +587,7 @@ int iommufd_hwpt_page_response(struct iommufd_ucmd *ucmd)
 {
 	struct iommu_hwpt_page_response *cmd = ucmd->cmd;
 	struct iommu_page_response resp = {};
+	struct iommufd_fault *ifault = NULL;
 	struct iommufd_fault *curr, *next;
 	struct iommufd_hw_pagetable *hwpt;
 	struct iommufd_device *idev;
@@ -547,6 +609,7 @@ int iommufd_hwpt_page_response(struct iommufd_ucmd *ucmd)
 		if (curr->dev != idev->dev || curr->fault.grpid != cmd->grpid)
 			continue;
 
+		ifault = curr;
 		if ((cmd->flags & IOMMU_PGFAULT_FLAGS_PASID_VALID) &&
 		    cmd->pasid != curr->fault.pasid)
 			break;
@@ -555,6 +618,15 @@ int iommufd_hwpt_page_response(struct iommufd_ucmd *ucmd)
 		    !(cmd->flags & IOMMU_PGFAULT_FLAGS_PASID_VALID))
 			break;
 
+		/*
+		 * The timer has expired if it was not pending. Leave the
+		 * response to the timer function.
+		 */
+		if (!iommufd_fault_timer_teardown(curr)) {
+			rc = -ETIMEDOUT;
+			break;
+		}
+
 		resp.version = IOMMU_PAGE_RESP_VERSION_1;
 		resp.pasid = cmd->pasid;
 		resp.grpid = cmd->grpid;
@@ -564,11 +636,11 @@ int iommufd_hwpt_page_response(struct iommufd_ucmd *ucmd)
 
 		rc = iommu_page_response(idev->dev, &resp);
 		list_del_init(&curr->item);
-		kfree(curr);
 		break;
 	}
 	mutex_unlock(&hwpt->fault->mutex);
 
+	iommufd_put_fault(ifault);
 	iommufd_put_object(&idev->obj);
 out_put_hwpt:
 	iommufd_put_object(&hwpt->obj);
-- 
2.34.1


  parent reply	other threads:[~2023-05-30  5:41 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-30  5:37 [RFC PATCHES 00/17] IOMMUFD: Deliver IO page faults to user space Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 01/17] iommu: Move iommu fault data to linux/iommu.h Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 02/17] iommu: Support asynchronous I/O page fault response Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 03/17] iommu: Add helper to set iopf handler for domain Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 04/17] iommu: Pass device parameter to iopf handler Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 05/17] iommu: Split IO page fault handling from SVA Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 06/17] iommu: Add iommu page fault cookie helpers Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 07/17] iommufd: Add iommu page fault data Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 08/17] iommufd: IO page fault delivery initialization and release Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 09/17] iommufd: Add iommufd hwpt iopf handler Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 10/17] iommufd: Add IOMMU_HWPT_ALLOC_FLAGS_USER_PASID_TABLE for hwpt_alloc Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 11/17] iommufd: Deliver fault messages to user space Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 12/17] iommufd: Add io page fault response support Lu Baolu
2023-05-30  5:37 ` Lu Baolu [this message]
2023-05-30  5:37 ` [RFC PATCHES 14/17] iommufd: Drain all pending faults when destroying hwpt Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 15/17] iommufd: Allow new hwpt_alloc flags Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 16/17] iommufd/selftest: Add IOPF feature for mock devices Lu Baolu
2023-05-30  5:37 ` [RFC PATCHES 17/17] iommufd/selftest: Cover iopf-capable nested hwpt Lu Baolu
2023-05-30 18:50 ` [RFC PATCHES 00/17] IOMMUFD: Deliver IO page faults to user space Nicolin Chen
2023-05-31  2:10   ` Baolu Lu
2023-05-31  4:12     ` Nicolin Chen
2023-06-25  6:30   ` Baolu Lu
2023-06-25 19:21     ` Nicolin Chen
2023-06-26  3:10       ` Baolu Lu
2023-06-26 18:02         ` Nicolin Chen
2023-06-26 18:33     ` Jason Gunthorpe
2023-06-28  2:00       ` Baolu Lu
2023-06-28 12:49         ` Jason Gunthorpe
2023-06-29  1:07           ` Baolu Lu
2023-05-31  0:33 ` Jason Gunthorpe
2023-05-31  3:17   ` Baolu Lu
2023-06-23  6:18   ` Baolu Lu
2023-06-23 13:50     ` Jason Gunthorpe
2023-06-16 11:32 ` Jean-Philippe Brucker
2023-06-19  3:35   ` Baolu Lu
2023-06-26  9:51     ` Jean-Philippe Brucker
2023-06-19 12:58   ` Jason Gunthorpe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230530053724.232765-14-baolu.lu@linux.intel.com \
    --to=baolu.lu@linux.intel.com \
    --cc=iommu@lists.linux.dev \
    --cc=jacob.jun.pan@linux.intel.com \
    --cc=jean-philippe@linaro.org \
    --cc=jgg@ziepe.ca \
    --cc=joro@8bytes.org \
    --cc=kevin.tian@intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=nicolinc@nvidia.com \
    --cc=robin.murphy@arm.com \
    --cc=virtualization@lists.linux-foundation.org \
    --cc=will@kernel.org \
    --cc=yi.l.liu@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox