patches.lists.linux.dev archive mirror
 help / color / mirror / Atom feed
From: Nicolin Chen <nicolinc@nvidia.com>
To: Jason Gunthorpe <jgg@nvidia.com>
Cc: <iommu@lists.linux.dev>, Joerg Roedel <joro@8bytes.org>,
	Kevin Tian <kevin.tian@intel.com>,
	<linux-kselftest@vger.kernel.org>,
	Robin Murphy <robin.murphy@arm.com>,
	Shuah Khan <shuah@kernel.org>, Will Deacon <will@kernel.org>,
	Lu Baolu <baolu.lu@linux.intel.com>, <patches@lists.linux.dev>,
	<syzbot+80620e2d0d0a33b09f93@syzkaller.appspotmail.com>
Subject: Re: [PATCH 1/3] iommufd: Fix race during abort for file descriptors
Date: Wed, 17 Sep 2025 22:07:38 -0700	[thread overview]
Message-ID: <aMuTmuBKd9aU7ngO@Asurada-Nvidia> (raw)
In-Reply-To: <1-v1-02cd136829df+31-iommufd_syz_fput_jgg@nvidia.com>

On Wed, Sep 17, 2025 at 05:01:47PM -0300, Jason Gunthorpe wrote:
> Fix this by putting the core code in charge of the file lifetime, and call
> __fput_sync() during abort to ensure that release() is called before
> kfree. __fput_sync() is a bit too tricky to open code in all the object
> implementations

Mind elaborating this "too tricky"? I thought that we're supposed
to use __fput_sync(), instead of fput(), in the alloc function in
the first place?

> Cc: stable@vger.kernel.org
> Fixes: 07838f7fd529 ("iommufd: Add iommufd fault object")
> Reported-by: syzbot+80620e2d0d0a33b09f93@syzkaller.appspotmail.com
> Closes: https://lore.kernel.org/r/68c8583d.050a0220.2ff435.03a2.GAE@google.com
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

The patch looks good to me though:

Reviewed-by: Nicolin Chen <nicolinc@nvidia.com>

> @@ -131,10 +132,30 @@ void iommufd_object_abort(struct iommufd_ctx *ictx, struct iommufd_object *obj)
>  void iommufd_object_abort_and_destroy(struct iommufd_ctx *ictx,
>  				      struct iommufd_object *obj)
>  {
> -	if (iommufd_object_ops[obj->type].abort)
> -		iommufd_object_ops[obj->type].abort(obj);
> +	const struct iommufd_object_ops *ops = &iommufd_object_ops[obj->type];
> +
> +	if (ops->file_offset) {
> +		struct file **filep = ((void *)obj) + ops->file_offset;
> +
> +		/*
> +		 * files should hold a users refcount while the file is open and
> +		 * put it back in their release. They should hold a pointer to
> +		 * obj in their private data. Normal fput() is deferred to a

Nit: there is only one file_offset per obj, so it should be "file"
and "it/its"?

> +		 * workqueue and can get out of order with the following
> +		 * kfree(obj). Using the sync version ensures the release
> +		 * happens immediately. During abort we require the file
> +		 * refcount is one at this point - meaning the object alloc
> +		 * function cannot do anything to allow another thread to take a
> +		 * refcount prior to a guaranteed success.
> +		 */

Thanks
Nicolin

  reply	other threads:[~2025-09-18  5:08 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-09-17 20:01 [PATCH 0/3] Fix a race with fput during eventq abort Jason Gunthorpe
2025-09-17 20:01 ` [PATCH 1/3] iommufd: Fix race during abort for file descriptors Jason Gunthorpe
2025-09-18  5:07   ` Nicolin Chen [this message]
2025-09-18 14:43     ` Jason Gunthorpe
2025-09-18 12:37   ` Nirmoy Das
2025-09-19  8:16   ` Tian, Kevin
2025-09-17 20:01 ` [PATCH 2/3] iommufd: WARN if an object is aborted with an elevated refcount Jason Gunthorpe
2025-09-18  6:10   ` Nicolin Chen
2025-09-18 14:47     ` Jason Gunthorpe
2025-09-18 20:49       ` Nicolin Chen
2025-09-18 20:50   ` Nicolin Chen
2025-09-19  8:16   ` Tian, Kevin
2025-09-17 20:01 ` [PATCH 3/3] iommufd/selftest: Update the fail_nth limit Jason Gunthorpe
2025-09-18  5:28   ` Nicolin Chen
2025-09-19  8:17   ` Tian, Kevin
2025-09-18 20:52 ` [PATCH 0/3] Fix a race with fput during eventq abort Nicolin Chen
2025-09-19 13:43 ` 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=aMuTmuBKd9aU7ngO@Asurada-Nvidia \
    --to=nicolinc@nvidia.com \
    --cc=baolu.lu@linux.intel.com \
    --cc=iommu@lists.linux.dev \
    --cc=jgg@nvidia.com \
    --cc=joro@8bytes.org \
    --cc=kevin.tian@intel.com \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=patches@lists.linux.dev \
    --cc=robin.murphy@arm.com \
    --cc=shuah@kernel.org \
    --cc=syzbot+80620e2d0d0a33b09f93@syzkaller.appspotmail.com \
    --cc=will@kernel.org \
    /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;
as well as URLs for NNTP newsgroup(s).