Linux IOMMU Development
 help / color / mirror / Atom feed
From: Joel Granados <j.granados@samsung.com>
To: Lu Baolu <baolu.lu@linux.intel.com>
Cc: 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>,
	<iommu@lists.linux.dev>, <linux-kselftest@vger.kernel.org>,
	<virtualization@lists.linux-foundation.org>,
	<linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2 4/6] iommufd: Deliver fault messages to user space
Date: Thu, 7 Dec 2023 17:34:10 +0100	[thread overview]
Message-ID: <20231207163410.ap3w4faii6wkgwed@localhost> (raw)
In-Reply-To: <20231026024930.382898-5-baolu.lu@linux.intel.com>

[-- Attachment #1: Type: text/plain, Size: 10080 bytes --]

On Thu, Oct 26, 2023 at 10:49:28AM +0800, Lu Baolu wrote:
> Add the file interface that provides a simple and efficient way for
> userspace to handle page faults. The file interface allows userspace
> to read fault messages sequentially, and to respond to the handling
> result by writing to the same file.
> 
> Userspace applications are recommended to use io_uring to speed up read
> and write efficiency.
> 
> With this done, allow userspace application to allocate a hw page table
> with IOMMU_HWPT_ALLOC_IOPF_CAPABLE flag set.
> 
> Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
> ---
>  drivers/iommu/iommufd/iommufd_private.h |   2 +
>  drivers/iommu/iommufd/hw_pagetable.c    | 204 +++++++++++++++++++++++-
>  2 files changed, 205 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/iommu/iommufd/iommufd_private.h b/drivers/iommu/iommufd/iommufd_private.h
> index 0dbaa2dc5b22..ff063bc48150 100644
> --- a/drivers/iommu/iommufd/iommufd_private.h
> +++ b/drivers/iommu/iommufd/iommufd_private.h
> @@ -237,6 +237,8 @@ struct hw_pgtable_fault {
>  	struct mutex mutex;
>  	struct list_head deliver;
>  	struct list_head response;
> +	struct file *fault_file;
> +	int fault_fd;
>  };
>  
>  /*
> diff --git a/drivers/iommu/iommufd/hw_pagetable.c b/drivers/iommu/iommufd/hw_pagetable.c
> index 9f94c824cf86..f0aac1bb2d2d 100644
> --- a/drivers/iommu/iommufd/hw_pagetable.c
> +++ b/drivers/iommu/iommufd/hw_pagetable.c
> @@ -3,6 +3,8 @@
>   * Copyright (c) 2021-2022, NVIDIA CORPORATION & AFFILIATES
>   */
>  #include <linux/iommu.h>
> +#include <linux/file.h>
> +#include <linux/anon_inodes.h>
>  #include <uapi/linux/iommufd.h>
>  
>  #include "../iommu-priv.h"
> @@ -38,9 +40,198 @@ static void iommufd_kernel_managed_hwpt_destroy(struct iommufd_object *obj)
>  	refcount_dec(&hwpt->ioas->obj.users);
>  }
>  
> +static int iommufd_compose_fault_message(struct iommu_fault *fault,
> +					 struct iommu_hwpt_pgfault *hwpt_fault,
> +					 struct device *dev)
> +{
> +	struct iommufd_device *idev = iopf_pasid_cookie_get(dev, IOMMU_NO_PASID);
> +
> +	if (!idev)
> +		return -ENODEV;
> +
> +	if (IS_ERR(idev))
> +		return PTR_ERR(idev);
> +
> +	hwpt_fault->size = sizeof(*hwpt_fault);
> +	hwpt_fault->flags = fault->prm.flags;
> +	hwpt_fault->dev_id = idev->obj.id;
> +	hwpt_fault->pasid = fault->prm.pasid;
> +	hwpt_fault->grpid = fault->prm.grpid;
> +	hwpt_fault->perm = fault->prm.perm;
> +	hwpt_fault->addr = fault->prm.addr;
> +	hwpt_fault->private_data[0] = fault->prm.private_data[0];
> +	hwpt_fault->private_data[1] = fault->prm.private_data[1];
> +
> +	return 0;
> +}
> +
> +static ssize_t hwpt_fault_fops_read(struct file *filep, char __user *buf,
> +				    size_t count, loff_t *ppos)
> +{
> +	size_t fault_size = sizeof(struct iommu_hwpt_pgfault);
> +	struct hw_pgtable_fault *fault = filep->private_data;
> +	struct iommu_hwpt_pgfault data;
> +	struct iopf_group *group;
> +	struct iopf_fault *iopf;
> +	size_t done = 0;
> +	int rc;
> +
> +	if (*ppos || count % fault_size)
> +		return -ESPIPE;
> +
> +	mutex_lock(&fault->mutex);
> +	while (!list_empty(&fault->deliver) && count > done) {
> +		group = list_first_entry(&fault->deliver,
> +					 struct iopf_group, node);
> +
> +		if (list_count_nodes(&group->faults) * fault_size > count - done)
> +			break;
> +
> +		list_for_each_entry(iopf, &group->faults, list) {
> +			rc = iommufd_compose_fault_message(&iopf->fault,
> +							   &data, group->dev);
> +			if (rc)
> +				goto err_unlock;
> +			rc = copy_to_user(buf + done, &data, fault_size);
> +			if (rc)
> +				goto err_unlock;
> +			done += fault_size;
> +		}
> +
> +		list_move_tail(&group->node, &fault->response);
> +	}
> +	mutex_unlock(&fault->mutex);
> +
> +	return done;
> +err_unlock:
> +	mutex_unlock(&fault->mutex);
> +	return rc;
> +}
> +
> +static ssize_t hwpt_fault_fops_write(struct file *filep,
> +				     const char __user *buf,
> +				     size_t count, loff_t *ppos)
> +{
> +	size_t response_size = sizeof(struct iommu_hwpt_page_response);
> +	struct hw_pgtable_fault *fault = filep->private_data;
> +	struct iommu_hwpt_page_response response;
> +	struct iommufd_hw_pagetable *hwpt;
> +	struct iopf_group *iter, *group;
> +	struct iommufd_device *idev;
> +	size_t done = 0;
> +	int rc = 0;
> +
> +	if (*ppos || count % response_size)
> +		return -ESPIPE;
> +
> +	mutex_lock(&fault->mutex);
> +	while (!list_empty(&fault->response) && count > done) {
> +		rc = copy_from_user(&response, buf + done, response_size);
> +		if (rc)
> +			break;
> +
> +		/* Get the device that this response targets at. */
> +		idev = container_of(iommufd_get_object(fault->ictx,
> +						       response.dev_id,
> +						       IOMMUFD_OBJ_DEVICE),
> +				    struct iommufd_device, obj);
> +		if (IS_ERR(idev)) {
> +			rc = PTR_ERR(idev);
> +			break;
> +		}
> +
> +		/*
> +		 * Get the hw page table that this response was generated for.
> +		 * It must match the one stored in the fault data.
> +		 */
> +		hwpt = container_of(iommufd_get_object(fault->ictx,
> +						       response.hwpt_id,
> +						       IOMMUFD_OBJ_HW_PAGETABLE),
> +				    struct iommufd_hw_pagetable, obj);
> +		if (IS_ERR(hwpt)) {
> +			iommufd_put_object(&idev->obj);
> +			rc = PTR_ERR(hwpt);
> +			break;
> +		}
> +
> +		if (hwpt != fault->hwpt) {
> +			rc = -EINVAL;
> +			goto put_obj;
> +		}
> +
> +		group = NULL;
> +		list_for_each_entry(iter, &fault->response, node) {
> +			if (response.grpid != iter->last_fault.fault.prm.grpid)
> +				continue;
> +
> +			if (idev->dev != iter->dev)
> +				continue;
> +
> +			if ((iter->last_fault.fault.prm.flags &
> +			     IOMMU_FAULT_PAGE_REQUEST_PASID_VALID) &&
> +			    response.pasid != iter->last_fault.fault.prm.pasid)
> +				continue;
> +
> +			group = iter;
> +			break;
> +		}
> +
> +		if (!group) {
> +			rc = -ENODEV;
> +			goto put_obj;
> +		}
> +
> +		rc = iopf_group_response(group, response.code);
> +		if (rc)
> +			goto put_obj;
> +
> +		list_del(&group->node);
> +		iopf_free_group(group);
> +		done += response_size;
> +put_obj:
> +		iommufd_put_object(&hwpt->obj);
> +		iommufd_put_object(&idev->obj);
> +		if (rc)
> +			break;
> +	}
> +	mutex_unlock(&fault->mutex);
> +
> +	return (rc < 0) ? rc : done;
> +}
> +
> +static const struct file_operations hwpt_fault_fops = {
> +	.owner		= THIS_MODULE,
> +	.read		= hwpt_fault_fops_read,
> +	.write		= hwpt_fault_fops_write,
> +};
> +
> +static int hw_pagetable_get_fault_fd(struct hw_pgtable_fault *fault)
> +{
> +	struct file *filep;
> +	int fdno;
> +
> +	fdno = get_unused_fd_flags(O_CLOEXEC);
> +	if (fdno < 0)
> +		return fdno;
> +
> +	filep = anon_inode_getfile("[iommufd-pgfault]", &hwpt_fault_fops,
> +				   fault, O_RDWR);
> +	if (IS_ERR(filep)) {
> +		put_unused_fd(fdno);
> +		return PTR_ERR(filep);
> +	}
> +
> +	fd_install(fdno, filep);
> +	fault->fault_file = filep;
> +	fault->fault_fd = fdno;
> +
> +	return 0;
> +}
> +
>  static struct hw_pgtable_fault *hw_pagetable_fault_alloc(void)
>  {
>  	struct hw_pgtable_fault *fault;
> +	int rc;
>  
>  	fault = kzalloc(sizeof(*fault), GFP_KERNEL);
>  	if (!fault)
> @@ -50,6 +241,12 @@ static struct hw_pgtable_fault *hw_pagetable_fault_alloc(void)
>  	INIT_LIST_HEAD(&fault->response);
>  	mutex_init(&fault->mutex);
>  
> +	rc = hw_pagetable_get_fault_fd(fault);
> +	if (rc) {
> +		kfree(fault);
> +		return ERR_PTR(rc);
> +	}
> +
>  	return fault;
>  }
>  
> @@ -58,6 +255,8 @@ static void hw_pagetable_fault_free(struct hw_pgtable_fault *fault)
>  	WARN_ON(!list_empty(&fault->deliver));
>  	WARN_ON(!list_empty(&fault->response));
>  
> +	fput(fault->fault_file);
> +	put_unused_fd(fault->fault_fd);
I have been running your code and have run into some invalid memory in
this line. When `put_unused_fd` is called the files of the current task
is accessed with `current->files`. In my case this is 0x0.

The reason for it being 0x0 is that `do_exit` calls `exit_files` where
the task files get set to NULL; this call is made in `do_exit` before we
execute `exit_task_work`.

'exit_task_work` is the call that eventually arrives here to `hw_pagetable_fault_free`.

The way I have arrived to this state is the following:
1. Version of linux kernel that I'm using : commit 357b5abcba0477f7f1391dd0fa3a919a6f06bdf0 (HEAD, lubaolu/iommufd-io-pgfault-delivery-v2)
2. Version of qemu that Im using : commit 577ef478780597d3f449feb01e853b93fa5c5530 (HEAD, yiliu/zhenzhong/wip/iommufd_nesting_rfcv1)
3. This error happens when my user space app is exiting. (hence the call
   to `do_exit`
4. I call the IOMMU_HWPT_ALLOC ioctl with
  .flags = IOMMU_HWPT_ALLOC_IOPF_CAPABLE and 
  .hwpt_type = IOMMU_HWPT_TYPE_DEFAULT
  .pt_id = the default ioas id.

I have resolved this in a naive way by just not calling the
put_unused_fd function.

Have you run into this? Is this a path that you were expecting?
Also, please get back to me if you need more information about how I got
to this place. I have provided what I think is enough info, but I might
be missing something obvious.

Best

>  	kfree(fault);
>  }
>  
> @@ -347,7 +546,9 @@ int iommufd_hwpt_alloc(struct iommufd_ucmd *ucmd)
>  	struct mutex *mutex;
>  	int rc;
>  
> -	if (cmd->flags & ~IOMMU_HWPT_ALLOC_NEST_PARENT || cmd->__reserved)
> +	if ((cmd->flags & ~(IOMMU_HWPT_ALLOC_NEST_PARENT |
> +			    IOMMU_HWPT_ALLOC_IOPF_CAPABLE)) ||
> +	    cmd->__reserved)
>  		return -EOPNOTSUPP;
>  	if (!cmd->data_len && cmd->hwpt_type != IOMMU_HWPT_TYPE_DEFAULT)
>  		return -EINVAL;
> @@ -416,6 +617,7 @@ int iommufd_hwpt_alloc(struct iommufd_ucmd *ucmd)
>  		hwpt->fault->hwpt = hwpt;
>  		hwpt->domain->iopf_handler = iommufd_hw_pagetable_iopf_handler;
>  		hwpt->domain->fault_data = hwpt;
> +		cmd->out_fault_fd = hwpt->fault->fault_fd;
>  	}
>  
>  	cmd->out_hwpt_id = hwpt->obj.id;
> -- 
> 2.34.1
> 

-- 

Joel Granados

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 659 bytes --]

  parent reply	other threads:[~2023-12-07 16:34 UTC|newest]

Thread overview: 46+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <CGME20231204150747eucas1p2365e92a7ac33ba99b801d7c800acaf6a@eucas1p2.samsung.com>
2023-10-26  2:49 ` [PATCH v2 0/6] IOMMUFD: Deliver IO page faults to user space Lu Baolu
2023-10-26  2:49   ` [PATCH v2 1/6] iommu: Add iommu page fault cookie helpers Lu Baolu
2023-12-01 14:38     ` Jason Gunthorpe
2023-12-08  6:24       ` Baolu Lu
2023-10-26  2:49   ` [PATCH v2 2/6] iommufd: Add iommu page fault uapi data Lu Baolu
2023-12-01 15:14     ` Jason Gunthorpe
2023-12-08  6:35       ` Baolu Lu
2023-10-26  2:49   ` [PATCH v2 3/6] iommufd: Initializing and releasing IO page fault data Lu Baolu
2023-12-12 13:10     ` Joel Granados
2023-12-12 14:12       ` Jason Gunthorpe
2023-12-13  2:04         ` Baolu Lu
2023-12-13  2:15           ` Tian, Kevin
2023-12-13 13:19             ` Jason Gunthorpe
2023-10-26  2:49   ` [PATCH v2 4/6] iommufd: Deliver fault messages to user space Lu Baolu
2023-12-01 15:24     ` Jason Gunthorpe
2023-12-08 11:43       ` Baolu Lu
2023-12-07 16:34     ` Joel Granados [this message]
2023-12-07 17:17       ` Jason Gunthorpe
2023-12-08  5:47         ` Baolu Lu
2023-12-08 13:41           ` Jason Gunthorpe
2024-01-12 17:46     ` Shameerali Kolothum Thodi
2024-01-15 16:47       ` Jason Gunthorpe
2024-01-15 17:44         ` Shameerali Kolothum Thodi
2024-01-15 17:58           ` Jason Gunthorpe
2023-10-26  2:49   ` [PATCH v2 5/6] iommufd/selftest: Add IOMMU_TEST_OP_TRIGGER_IOPF test support Lu Baolu
2023-10-26  2:49   ` [PATCH v2 6/6] iommufd/selftest: Add coverage for IOMMU_TEST_OP_TRIGGER_IOPF Lu Baolu
2023-11-02 12:47   ` [PATCH v2 0/6] IOMMUFD: Deliver IO page faults to user space Jason Gunthorpe
2023-11-07  8:35     ` Tian, Kevin
2023-11-07 17:54       ` Jason Gunthorpe
2023-11-08  8:53         ` Tian, Kevin
2023-11-08 17:39           ` Jason Gunthorpe
     [not found]     ` <c774e157-9b47-4fb8-80dd-37441c69b43d@linux.intel.com>
2023-11-15 13:58       ` Jason Gunthorpe
2023-11-16  1:42         ` Liu, Jing2
2023-11-21  0:14           ` Jason Gunthorpe
2023-11-29  9:08   ` Shameerali Kolothum Thodi
2023-11-30  3:44     ` Baolu Lu
2023-12-01 14:24   ` Jason Gunthorpe
2023-12-08  5:57     ` Baolu Lu
2023-12-08 13:43       ` Jason Gunthorpe
2023-12-04 15:07   ` Joel Granados
2023-12-04 15:32     ` Jason Gunthorpe
2023-12-08  5:10     ` Baolu Lu
2024-01-12 21:56   ` Joel Granados
2024-01-14 13:13     ` Baolu Lu
2024-01-14 17:18       ` Joel Granados
2024-01-15  1:25         ` Baolu Lu

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=20231207163410.ap3w4faii6wkgwed@localhost \
    --to=j.granados@samsung.com \
    --cc=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