From: Yi Liu <yi.l.liu@intel.com> To: alex.williamson@redhat.com, jgg@nvidia.com, kevin.tian@intel.com Cc: cohuck@redhat.com, eric.auger@redhat.com, nicolinc@nvidia.com, kvm@vger.kernel.org, mjrosato@linux.ibm.com, chao.p.peng@linux.intel.com, yi.l.liu@intel.com, yi.y.sun@linux.intel.com, peterx@redhat.com, jasowang@redhat.com, shameerali.kolothum.thodi@huawei.com, lulu@redhat.com, suravee.suthikulpanit@amd.com, intel-gvt-dev@lists.freedesktop.org, intel-gfx@lists.freedesktop.org, linux-kernel@vger.kernel.org, linux-s390@vger.kernel.org Subject: [PATCH v2 07/14] vfio: Block device access via device fd until device is opened Date: Mon, 6 Feb 2023 01:05:25 -0800 [thread overview] Message-ID: <20230206090532.95598-8-yi.l.liu@intel.com> (raw) In-Reply-To: <20230206090532.95598-1-yi.l.liu@intel.com> Allow the vfio_device file to be in a state where the device FD is opened but the device cannot be used by userspace (i.e. its .open_device() hasn't been called). This inbetween state is not used when the device FD is spawned from the group FD, however when we create the device FD directly by opening a cdev it will be opened in the blocked state. The reason for the inbetween state is userspace only gets a FD but doesn't have the secure until binding the FD to an iommufd. So in the blocked state, only the bind operation is allowed, other device accesses are not allowed. Completing bind will allow user to further access the device. This is implemented by adding a flag in struct vfio_device_file to mark the blocked state and using a simple smp_load_acquire() to obtain the flag value and serialize all the device setup with the thread accessing this device. Following this lockless scheme, it can safely handle the device FD unbound->bound but it cannot handle bound->unbound. To allow this we'd need to add a lock on all the vfio ioctls which seems costly. So once device FD is bound, it remains bound until the FD is closed. Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> --- drivers/vfio/vfio.h | 1 + drivers/vfio/vfio_main.c | 34 +++++++++++++++++++++++++++++++++- 2 files changed, 34 insertions(+), 1 deletion(-) diff --git a/drivers/vfio/vfio.h b/drivers/vfio/vfio.h index d8275881c1f1..802e13f1256e 100644 --- a/drivers/vfio/vfio.h +++ b/drivers/vfio/vfio.h @@ -18,6 +18,7 @@ struct vfio_container; struct vfio_device_file { struct vfio_device *device; + bool access_granted; spinlock_t kvm_ref_lock; /* protect kvm field */ struct kvm *kvm; struct iommufd_ctx *iommufd; /* protected by struct vfio_device_set::lock */ diff --git a/drivers/vfio/vfio_main.c b/drivers/vfio/vfio_main.c index c517252aba19..2267057240bd 100644 --- a/drivers/vfio/vfio_main.c +++ b/drivers/vfio/vfio_main.c @@ -476,7 +476,15 @@ int vfio_device_open(struct vfio_device_file *df) device->open_count--; } - return ret; + if (ret) + return ret; + + /* + * Paired with smp_load_acquire() in vfio_device_fops::ioctl/ + * read/write/mmap + */ + smp_store_release(&df->access_granted, true); + return 0; } void vfio_device_close(struct vfio_device_file *df) @@ -1104,8 +1112,14 @@ static long vfio_device_fops_unl_ioctl(struct file *filep, { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; int ret; + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; + ret = vfio_device_pm_runtime_get(device); if (ret) return ret; @@ -1132,6 +1146,12 @@ static ssize_t vfio_device_fops_read(struct file *filep, char __user *buf, { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; + + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; if (unlikely(!device->ops->read)) return -EINVAL; @@ -1145,6 +1165,12 @@ static ssize_t vfio_device_fops_write(struct file *filep, { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; + + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; if (unlikely(!device->ops->write)) return -EINVAL; @@ -1156,6 +1182,12 @@ static int vfio_device_fops_mmap(struct file *filep, struct vm_area_struct *vma) { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; + + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; if (unlikely(!device->ops->mmap)) return -EINVAL; -- 2.34.1
WARNING: multiple messages have this Message-ID (diff)
From: Yi Liu <yi.l.liu@intel.com> To: alex.williamson@redhat.com, jgg@nvidia.com, kevin.tian@intel.com Cc: linux-s390@vger.kernel.org, yi.l.liu@intel.com, yi.y.sun@linux.intel.com, kvm@vger.kernel.org, mjrosato@linux.ibm.com, jasowang@redhat.com, cohuck@redhat.com, linux-kernel@vger.kernel.org, peterx@redhat.com, eric.auger@redhat.com, nicolinc@nvidia.com, shameerali.kolothum.thodi@huawei.com, suravee.suthikulpanit@amd.com, chao.p.peng@linux.intel.com, lulu@redhat.com, intel-gvt-dev@lists.freedesktop.org, intel-gfx@lists.freedesktop.org Subject: [Intel-gfx] [PATCH v2 07/14] vfio: Block device access via device fd until device is opened Date: Mon, 6 Feb 2023 01:05:25 -0800 [thread overview] Message-ID: <20230206090532.95598-8-yi.l.liu@intel.com> (raw) In-Reply-To: <20230206090532.95598-1-yi.l.liu@intel.com> Allow the vfio_device file to be in a state where the device FD is opened but the device cannot be used by userspace (i.e. its .open_device() hasn't been called). This inbetween state is not used when the device FD is spawned from the group FD, however when we create the device FD directly by opening a cdev it will be opened in the blocked state. The reason for the inbetween state is userspace only gets a FD but doesn't have the secure until binding the FD to an iommufd. So in the blocked state, only the bind operation is allowed, other device accesses are not allowed. Completing bind will allow user to further access the device. This is implemented by adding a flag in struct vfio_device_file to mark the blocked state and using a simple smp_load_acquire() to obtain the flag value and serialize all the device setup with the thread accessing this device. Following this lockless scheme, it can safely handle the device FD unbound->bound but it cannot handle bound->unbound. To allow this we'd need to add a lock on all the vfio ioctls which seems costly. So once device FD is bound, it remains bound until the FD is closed. Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> --- drivers/vfio/vfio.h | 1 + drivers/vfio/vfio_main.c | 34 +++++++++++++++++++++++++++++++++- 2 files changed, 34 insertions(+), 1 deletion(-) diff --git a/drivers/vfio/vfio.h b/drivers/vfio/vfio.h index d8275881c1f1..802e13f1256e 100644 --- a/drivers/vfio/vfio.h +++ b/drivers/vfio/vfio.h @@ -18,6 +18,7 @@ struct vfio_container; struct vfio_device_file { struct vfio_device *device; + bool access_granted; spinlock_t kvm_ref_lock; /* protect kvm field */ struct kvm *kvm; struct iommufd_ctx *iommufd; /* protected by struct vfio_device_set::lock */ diff --git a/drivers/vfio/vfio_main.c b/drivers/vfio/vfio_main.c index c517252aba19..2267057240bd 100644 --- a/drivers/vfio/vfio_main.c +++ b/drivers/vfio/vfio_main.c @@ -476,7 +476,15 @@ int vfio_device_open(struct vfio_device_file *df) device->open_count--; } - return ret; + if (ret) + return ret; + + /* + * Paired with smp_load_acquire() in vfio_device_fops::ioctl/ + * read/write/mmap + */ + smp_store_release(&df->access_granted, true); + return 0; } void vfio_device_close(struct vfio_device_file *df) @@ -1104,8 +1112,14 @@ static long vfio_device_fops_unl_ioctl(struct file *filep, { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; int ret; + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; + ret = vfio_device_pm_runtime_get(device); if (ret) return ret; @@ -1132,6 +1146,12 @@ static ssize_t vfio_device_fops_read(struct file *filep, char __user *buf, { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; + + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; if (unlikely(!device->ops->read)) return -EINVAL; @@ -1145,6 +1165,12 @@ static ssize_t vfio_device_fops_write(struct file *filep, { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; + + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; if (unlikely(!device->ops->write)) return -EINVAL; @@ -1156,6 +1182,12 @@ static int vfio_device_fops_mmap(struct file *filep, struct vm_area_struct *vma) { struct vfio_device_file *df = filep->private_data; struct vfio_device *device = df->device; + bool access; + + /* Paired with smp_store_release() in vfio_device_open() */ + access = smp_load_acquire(&df->access_granted); + if (!access) + return -EINVAL; if (unlikely(!device->ops->mmap)) return -EINVAL; -- 2.34.1
next prev parent reply other threads:[~2023-02-06 9:06 UTC|newest] Thread overview: 63+ messages / expand[flat|nested] mbox.gz Atom feed top 2023-02-06 9:05 [PATCH v2 00/14] Add vfio_device cdev for iommufd support Yi Liu 2023-02-06 9:05 ` [Intel-gfx] " Yi Liu 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 01/14] vfio: Allocate per device file structure Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 02/14] vfio: Refine vfio file kAPIs Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-06 9:05 ` [PATCH v2 03/14] vfio: Accept vfio device file in the driver facing kAPI Yi Liu 2023-02-06 9:05 ` [Intel-gfx] " Yi Liu 2023-02-07 3:25 ` Tian, Kevin 2023-02-07 3:25 ` [Intel-gfx] " Tian, Kevin 2023-02-06 9:05 ` [PATCH v2 04/14] kvm/vfio: Rename kvm_vfio_group to prepare for accepting vfio device fd Yi Liu 2023-02-06 9:05 ` [Intel-gfx] " Yi Liu 2023-02-06 9:05 ` [PATCH v2 05/14] kvm/vfio: Accept vfio device file from userspace Yi Liu 2023-02-06 9:05 ` [Intel-gfx] " Yi Liu 2023-02-07 3:36 ` Tian, Kevin 2023-02-07 3:36 ` [Intel-gfx] " Tian, Kevin 2023-02-07 8:39 ` Liu, Yi L 2023-02-07 8:39 ` [Intel-gfx] " Liu, Yi L 2023-02-06 9:05 ` [PATCH v2 06/14] vfio: Pass struct vfio_device_file * to vfio_device_open/close() Yi Liu 2023-02-06 9:05 ` [Intel-gfx] " Yi Liu 2023-02-07 3:42 ` Tian, Kevin 2023-02-07 3:42 ` [Intel-gfx] " Tian, Kevin 2023-02-06 9:05 ` Yi Liu [this message] 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 07/14] vfio: Block device access via device fd until device is opened Yi Liu 2023-02-07 3:52 ` Tian, Kevin 2023-02-07 3:52 ` [Intel-gfx] " Tian, Kevin 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 08/14] vfio: Add infrastructure for bind_iommufd from userspace Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-07 3:56 ` [Intel-gfx] " Tian, Kevin 2023-02-07 3:56 ` Tian, Kevin 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 09/14] vfio-iommufd: Add detach_ioas support for physical VFIO devices Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-07 6:06 ` Tian, Kevin 2023-02-07 6:06 ` [Intel-gfx] " Tian, Kevin 2023-02-07 8:56 ` Liu, Yi L 2023-02-07 8:56 ` [Intel-gfx] " Liu, Yi L 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 10/14] vfio-iommufd: Add detach_ioas for emulated " Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-07 6:08 ` Tian, Kevin 2023-02-07 6:08 ` [Intel-gfx] " Tian, Kevin 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 11/14] vfio: Make vfio_device_open() exclusive between group path and device cdev path Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-07 6:24 ` Tian, Kevin 2023-02-07 6:24 ` [Intel-gfx] " Tian, Kevin 2023-02-07 8:54 ` Liu, Yi L 2023-02-07 8:54 ` [Intel-gfx] " Liu, Yi L 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 12/14] vfio: Add cdev for vfio_device Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-07 7:54 ` Tian, Kevin 2023-02-07 7:54 ` [Intel-gfx] " Tian, Kevin 2023-02-10 11:31 ` Joao Martins 2023-02-10 11:31 ` [Intel-gfx] " Joao Martins 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 13/14] vfio: Add ioctls for device cdev using iommufd Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-07 3:41 ` Yan Zhao 2023-02-07 3:41 ` [Intel-gfx] " Yan Zhao 2023-02-07 7:50 ` Liu, Yi L 2023-02-07 7:50 ` Liu, Yi L 2023-02-07 9:17 ` Tian, Kevin 2023-02-07 9:17 ` [Intel-gfx] " Tian, Kevin 2023-02-06 9:05 ` [Intel-gfx] [PATCH v2 14/14] vfio: Compile group optionally Yi Liu 2023-02-06 9:05 ` Yi Liu 2023-02-06 11:55 ` [Intel-gfx] ✗ Fi.CI.BUILD: failure for Add vfio_device cdev for iommufd support Patchwork
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=20230206090532.95598-8-yi.l.liu@intel.com \ --to=yi.l.liu@intel.com \ --cc=alex.williamson@redhat.com \ --cc=chao.p.peng@linux.intel.com \ --cc=cohuck@redhat.com \ --cc=eric.auger@redhat.com \ --cc=intel-gfx@lists.freedesktop.org \ --cc=intel-gvt-dev@lists.freedesktop.org \ --cc=jasowang@redhat.com \ --cc=jgg@nvidia.com \ --cc=kevin.tian@intel.com \ --cc=kvm@vger.kernel.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-s390@vger.kernel.org \ --cc=lulu@redhat.com \ --cc=mjrosato@linux.ibm.com \ --cc=nicolinc@nvidia.com \ --cc=peterx@redhat.com \ --cc=shameerali.kolothum.thodi@huawei.com \ --cc=suravee.suthikulpanit@amd.com \ --cc=yi.y.sun@linux.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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.