linux-kernel - Re: [PATCH 2/4] panthor: save panthor_file in panthor

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250623082122.62f69579@fedora>
Date: Mon, 23 Jun 2025 08:21:22 +0200
From: Boris Brezillon <boris.brezillon@...labora.com>
To: Chia-I Wu <olvaffe@...il.com>
Cc: Steven Price <steven.price@....com>, Liviu Dudau <liviu.dudau@....com>,
 Maarten Lankhorst <maarten.lankhorst@...ux.intel.com>, Maxime Ripard
 <mripard@...nel.org>, Thomas Zimmermann <tzimmermann@...e.de>, David Airlie
 <airlied@...il.com>, dri-devel@...ts.freedesktop.org,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/4] panthor: save panthor_file in panthor_group

On Fri, 20 Jun 2025 16:50:51 -0700
Chia-I Wu <olvaffe@...il.com> wrote:

> We would like to access panthor_file from panthor_group on gpu errors.
> Because panthour_group can outlive drm_file, add refcount to
> panthor_file to ensure its lifetime.

I'm not a huge fan of refcounting panthor_file because people tend to
put resource they expect to be released when the last handle goes away,
and if we don't refcount these sub-resources they might live longer
than they are meant to. Also not a huge fan of the circular referencing
that exists between file and groups after this change.

How about we move the process info to a sub-object that's refcounted
and let both panthor_file and panthor_group take a ref on this object
instead?

> 
> Signed-off-by: Chia-I Wu <olvaffe@...il.com>
> ---
>  drivers/gpu/drm/panthor/panthor_device.h | 16 ++++++++++++++++
>  drivers/gpu/drm/panthor/panthor_drv.c    | 15 ++++++++++++++-
>  drivers/gpu/drm/panthor/panthor_mmu.c    |  1 +
>  drivers/gpu/drm/panthor/panthor_sched.c  |  6 ++++++
>  4 files changed, 37 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_device.h b/drivers/gpu/drm/panthor/panthor_device.h
> index 4fc7cf2aeed57..75ae6fd3a5128 100644
> --- a/drivers/gpu/drm/panthor/panthor_device.h
> +++ b/drivers/gpu/drm/panthor/panthor_device.h
> @@ -256,8 +256,24 @@ struct panthor_file {
>  
>  	/** @stats: cycle and timestamp measures for job execution. */
>  	struct panthor_gpu_usage stats;
> +
> +	/** @refcount: ref count of this file */
> +	struct kref refcount;
>  };
>  
> +static inline struct panthor_file *panthor_file_get(struct panthor_file *pfile)
> +{
> +	kref_get(&pfile->refcount);
> +	return pfile;
> +}
> +
> +void panthor_file_release(struct kref *kref);
> +
> +static inline void panthor_file_put(struct panthor_file *pfile)
> +{
> +	kref_put(&pfile->refcount, panthor_file_release);
> +}
> +
>  int panthor_device_init(struct panthor_device *ptdev);
>  void panthor_device_unplug(struct panthor_device *ptdev);
>  
> diff --git a/drivers/gpu/drm/panthor/panthor_drv.c b/drivers/gpu/drm/panthor/panthor_drv.c
> index 775a66c394544..aea9609684b77 100644
> --- a/drivers/gpu/drm/panthor/panthor_drv.c
> +++ b/drivers/gpu/drm/panthor/panthor_drv.c
> @@ -1393,6 +1393,16 @@ static int panthor_ioctl_set_user_mmio_offset(struct drm_device *ddev,
>  	return 0;
>  }
>  
> +void panthor_file_release(struct kref *kref)
> +{
> +	struct panthor_file *pfile =
> +		container_of(kref, struct panthor_file, refcount);
> +
> +	WARN_ON(pfile->vms || pfile->groups);
> +
> +	kfree(pfile);
> +}
> +
>  static int
>  panthor_open(struct drm_device *ddev, struct drm_file *file)
>  {
> @@ -1426,6 +1436,8 @@ panthor_open(struct drm_device *ddev, struct drm_file *file)
>  	if (ret)
>  		goto err_destroy_vm_pool;
>  
> +	kref_init(&pfile->refcount);
> +
>  	file->driver_priv = pfile;
>  	return 0;
>  
> @@ -1442,10 +1454,11 @@ panthor_postclose(struct drm_device *ddev, struct drm_file *file)
>  {
>  	struct panthor_file *pfile = file->driver_priv;
>  
> +	/* destroy vm and group handles now to avoid circular references */
>  	panthor_group_pool_destroy(pfile);
>  	panthor_vm_pool_destroy(pfile);
>  
> -	kfree(pfile);
> +	panthor_file_put(pfile);
>  }
>  
>  static const struct drm_ioctl_desc panthor_drm_driver_ioctls[] = {
> diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
> index b39ea6acc6a96..ccbcfe11420ac 100644
> --- a/drivers/gpu/drm/panthor/panthor_mmu.c
> +++ b/drivers/gpu/drm/panthor/panthor_mmu.c
> @@ -1604,6 +1604,7 @@ void panthor_vm_pool_destroy(struct panthor_file *pfile)
>  
>  	xa_destroy(&pfile->vms->xa);
>  	kfree(pfile->vms);
> +	pfile->vms = NULL;
>  }
>  
>  /**
> diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
> index a2248f692a030..485072904cd7d 100644
> --- a/drivers/gpu/drm/panthor/panthor_sched.c
> +++ b/drivers/gpu/drm/panthor/panthor_sched.c
> @@ -535,6 +535,9 @@ struct panthor_group {
>  	/** @ptdev: Device. */
>  	struct panthor_device *ptdev;
>  
> +	/** @pfile: File this group is created from. */
> +	struct panthor_file *pfile;
> +
>  	/** @vm: VM bound to the group. */
>  	struct panthor_vm *vm;
>  
> @@ -919,6 +922,7 @@ static void group_release_work(struct work_struct *work)
>  	panthor_kernel_bo_destroy(group->syncobjs);
>  
>  	panthor_vm_put(group->vm);
> +	panthor_file_put(group->pfile);
>  	kfree(group);
>  }
>  
> @@ -3467,6 +3471,8 @@ int panthor_group_create(struct panthor_file *pfile,
>  	INIT_WORK(&group->tiler_oom_work, group_tiler_oom_work);
>  	INIT_WORK(&group->release_work, group_release_work);
>  
> +	group->pfile = panthor_file_get(pfile);
> +
>  	group->vm = panthor_vm_pool_get_vm(pfile->vms, group_args->vm_id);
>  	if (!group->vm) {
>  		ret = -EINVAL;