public inbox for intel-xe@lists.freedesktop.org
 help / color / mirror / Atom feed
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
To: Satyanarayana K V P <satyanarayana.k.v.p@intel.com>
Cc: intel-xe@lists.freedesktop.org,
	"Michal Wajdeczko" <michal.wajdeczko@intel.com>,
	"Piotr Piórkowski" <piotr.piorkowski@intel.com>,
	"Matthew Brost" <matthew.brost@intel.com>,
	"Thomas Hellström" <thomas.hellstrom@linux.intel.com>,
	"Michał Winiarski" <michal.winiarski@intel.com>,
	"Dunajski Bartosz" <bartosz.dunajski@intel.com>,
	dri-devel@lists.freedesktop.org
Subject: Re: [RFC v6 1/1] drm/xe/pf: Restrict device query responses in admin-only PF mode
Date: Mon, 23 Mar 2026 18:03:01 -0400	[thread overview]
Message-ID: <acG4lRQGh9ZCVZDW@intel.com> (raw)
In-Reply-To: <20260316064100.2542412-4-satyanarayana.k.v.p@intel.com>

On Mon, Mar 16, 2026 at 06:41:02AM +0000, Satyanarayana K V P wrote:
> When a PF is configured in admin-only mode, it is intended for management
> only and must not expose workload-facing capabilities to userspace.
> 
> Limit the exposed ioctl set in admin-only PF mode to XE_DEVICE_QUERY, and
> suppress capability-bearing query payloads so that the userspace cannot
> discover execution-related device details in this mode.
> 
> Enable admin-only mode with:
> echo 0 | sudo tee /sys/bus/pci/drivers_autoprobe
> sudo modprobe xe

instead of playing with autoprobe/modprobe we can assume it is already
probed and then we can simply put the
echo $BDF | sudo tee /sys/bus/pci/drivers/xe/unbind

> sudo mkdir /sys/kernel/config/xe/<B:D:F>
> echo yes | sudo tee /sys/kernel/config/xe/<B:D:F>/sriov/admin_only_pf
> echo <B:D:F> | sudo tee /sys/bus/pci/drivers/xe/bind
> 
> Signed-off-by: Satyanarayana K V P <satyanarayana.k.v.p@intel.com>
> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Cc: Piotr Piórkowski <piotr.piorkowski@intel.com>
> Cc: Matthew Brost <matthew.brost@intel.com>
> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: Michał Winiarski <michal.winiarski@intel.com>
> Cc: Dunajski Bartosz <bartosz.dunajski@intel.com>
> Cc: dri-devel@lists.freedesktop.org
> 
> ---
> V5 -> V6:
> - Updated commit message.
> - Return number of engines and memory regions as zero instead of returning
> query size as zero (Michal Wajdeczko).
> - Allow all other query IOCTLs excepts query_engines and query_mem_regions
> (Michal Wajdeczko).

Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

If Mesa gets ready with this and Compute folks also agree I believe we
should already take this patch in.

> 
> V4 -> V5:
> - Updated commit message (Matt B).
> - Introduced new driver_admin_only_pf structure (Michal Wajdeczko).
> - Updated all query configs (Michal Wajdeczko).
> - Renamed xe_device_is_admin_only() to xe_device_is_admin_only_pf()
> - Fixed other review comments (Michal Wajdeczko).
> 
> V3 -> V4:
> - Suppressed device capabilities in admin-only PF mode. (Wajdeczko)
> 
> V2 -> V3:
> - Introduced new helper function xe_debugfs_create_files() to create
> debugfs entries based on admin_only_pf mode or normal mode.
> 
> V1 -> V2:
> - Rebased to latest drm-tip.
> - Update update_minor_dev() to debugfs_minor_dev().
> ---
>  drivers/gpu/drm/xe/xe_device.c | 60 +++++++++++++++++++++++++++++++---
>  drivers/gpu/drm/xe/xe_device.h |  1 +
>  drivers/gpu/drm/xe/xe_query.c  |  6 ++++
>  3 files changed, 63 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
> index e77a3a3db73d..7decc6510681 100644
> --- a/drivers/gpu/drm/xe/xe_device.c
> +++ b/drivers/gpu/drm/xe/xe_device.c
> @@ -25,6 +25,7 @@
>  #include "regs/xe_regs.h"
>  #include "xe_bo.h"
>  #include "xe_bo_evict.h"
> +#include "xe_configfs.h"
>  #include "xe_debugfs.h"
>  #include "xe_defaults.h"
>  #include "xe_devcoredump.h"
> @@ -213,6 +214,10 @@ static const struct drm_ioctl_desc xe_ioctls[] = {
>  			  DRM_RENDER_ALLOW),
>  };
>  
> +static const struct drm_ioctl_desc xe_ioctls_admin_only[] = {
> +	DRM_IOCTL_DEF_DRV(XE_DEVICE_QUERY, xe_query_ioctl, DRM_RENDER_ALLOW),
> +};
> +
>  static long xe_drm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
>  {
>  	struct drm_file *file_priv = file->private_data;
> @@ -387,7 +392,7 @@ bool xe_is_xe_file(const struct file *file)
>  	return file->f_op == &xe_driver_fops;
>  }
>  
> -static struct drm_driver driver = {
> +static struct drm_driver regular_driver = {
>  	/* Don't use MTRRs here; the Xserver or userspace app should
>  	 * deal with them for Intel hardware.
>  	 */
> @@ -415,6 +420,40 @@ static struct drm_driver driver = {
>  	.patchlevel = DRIVER_PATCHLEVEL,
>  };
>  
> +static struct drm_driver admin_only_driver = {
> +	.driver_features =
> +	    DRIVER_GEM | DRIVER_RENDER | DRIVER_GEM_GPUVA,
> +	.open = xe_file_open,
> +	.postclose = xe_file_close,
> +
> +	.gem_prime_import = xe_gem_prime_import,
> +
> +	.dumb_create = xe_bo_dumb_create,
> +	.dumb_map_offset = drm_gem_ttm_dumb_map_offset,
> +#ifdef CONFIG_PROC_FS
> +	.show_fdinfo = xe_drm_client_fdinfo,
> +#endif
> +	.ioctls = xe_ioctls_admin_only,
> +	.num_ioctls = ARRAY_SIZE(xe_ioctls_admin_only),
> +	.fops = &xe_driver_fops,
> +	.name = DRIVER_NAME,
> +	.desc = DRIVER_DESC,
> +	.major = DRIVER_MAJOR,
> +	.minor = DRIVER_MINOR,
> +	.patchlevel = DRIVER_PATCHLEVEL,
> +};
> +
> +/**
> + * xe_device_is_admin_only() - Check whether device is admin only or not.
> + * @xe: the &xe_device to check
> + *
> + * Return: true if the device is admin only, false otherwise.
> + */
> +bool xe_device_is_admin_only(const struct xe_device *xe)
> +{
> +	return xe->drm.driver == &admin_only_driver;
> +}
> +
>  static void xe_device_destroy(struct drm_device *dev, void *dummy)
>  {
>  	struct xe_device *xe = to_xe_device(dev);
> @@ -439,16 +478,24 @@ static void xe_device_destroy(struct drm_device *dev, void *dummy)
>  struct xe_device *xe_device_create(struct pci_dev *pdev,
>  				   const struct pci_device_id *ent)
>  {
> +	struct drm_driver *driver = &regular_driver;
>  	struct xe_device *xe;
>  	int err;
>  
> -	xe_display_driver_set_hooks(&driver);
> +	/*
> +	 * Since XE device is not initialized yet, read from configfs
> +	 * directly to decide whether we are in admin-only PF mode or not.
> +	 */
> +	if (xe_configfs_admin_only_pf(pdev))
> +		driver = &admin_only_driver;
> +
> +	xe_display_driver_set_hooks(driver);
>  
> -	err = aperture_remove_conflicting_pci_devices(pdev, driver.name);
> +	err = aperture_remove_conflicting_pci_devices(pdev, driver->name);
>  	if (err)
>  		return ERR_PTR(err);
>  
> -	xe = devm_drm_dev_alloc(&pdev->dev, &driver, struct xe_device, drm);
> +	xe = devm_drm_dev_alloc(&pdev->dev, driver, struct xe_device, drm);
>  	if (IS_ERR(xe))
>  		return xe;
>  
> @@ -708,6 +755,11 @@ int xe_device_probe_early(struct xe_device *xe)
>  
>  	xe_sriov_probe_early(xe);
>  
> +	if (xe_configfs_admin_only_pf(to_pci_dev(xe->drm.dev)) && !IS_SRIOV_PF(xe)) {
> +		drm_err(&xe->drm, "Admin-only PF mode is enabled in non PF mode\n");
> +		return -ENODEV;
> +	}
> +
>  	if (IS_SRIOV_VF(xe))
>  		vf_update_device_info(xe);
>  
> diff --git a/drivers/gpu/drm/xe/xe_device.h b/drivers/gpu/drm/xe/xe_device.h
> index c4d267002661..4695761585e6 100644
> --- a/drivers/gpu/drm/xe/xe_device.h
> +++ b/drivers/gpu/drm/xe/xe_device.h
> @@ -43,6 +43,7 @@ static inline struct xe_device *ttm_to_xe_device(struct ttm_device *ttm)
>  	return container_of(ttm, struct xe_device, ttm);
>  }
>  
> +bool xe_device_is_admin_only(const struct xe_device *xe);
>  struct xe_device *xe_device_create(struct pci_dev *pdev,
>  				   const struct pci_device_id *ent);
>  int xe_device_probe_early(struct xe_device *xe);
> diff --git a/drivers/gpu/drm/xe/xe_query.c b/drivers/gpu/drm/xe/xe_query.c
> index 4852fdcb4b95..a42b6606a55b 100644
> --- a/drivers/gpu/drm/xe/xe_query.c
> +++ b/drivers/gpu/drm/xe/xe_query.c
> @@ -217,6 +217,9 @@ static int query_engines(struct xe_device *xe,
>  
>  	engines->num_engines = i;
>  
> +	if (xe_device_is_admin_only(xe))
> +		memset(engines, 0, size);
> +
>  	if (copy_to_user(query_ptr, engines, size)) {
>  		kfree(engines);
>  		return -EFAULT;
> @@ -297,6 +300,9 @@ static int query_mem_regions(struct xe_device *xe,
>  		}
>  	}
>  
> +	if (xe_device_is_admin_only(xe))
> +		memset(mem_regions, 0, size);
> +
>  	if (!copy_to_user(query_ptr, mem_regions, size))
>  		ret = 0;
>  	else
> -- 
> 2.43.0
> 

  reply	other threads:[~2026-03-23 22:03 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-16  6:41 [RFC v6 0/1] Do not create drm device for PF only admin mode Satyanarayana K V P
2026-03-16  6:41 ` [RFC v6 1/1] drm/xe/pf: Restrict device query responses in admin-only PF mode Satyanarayana K V P
2026-03-23 22:03   ` Rodrigo Vivi [this message]
2026-03-24 21:17   ` Dixit, Ashutosh
2026-03-25  5:21     ` K V P, Satyanarayana
2026-03-25 13:11       ` Rodrigo Vivi
2026-03-25  8:38     ` Michal Wajdeczko
2026-03-27  5:34       ` Dixit, Ashutosh
2026-03-27 13:26         ` Rodrigo Vivi
2026-03-16  6:47 ` ✓ CI.KUnit: success for Do not create drm device for PF only admin mode (rev5) Patchwork
2026-03-16  7:27 ` ✓ Xe.CI.BAT: " Patchwork
2026-03-17  8:16 ` ✗ Xe.CI.FULL: failure " 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=acG4lRQGh9ZCVZDW@intel.com \
    --to=rodrigo.vivi@intel.com \
    --cc=bartosz.dunajski@intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=matthew.brost@intel.com \
    --cc=michal.wajdeczko@intel.com \
    --cc=michal.winiarski@intel.com \
    --cc=piotr.piorkowski@intel.com \
    --cc=satyanarayana.k.v.p@intel.com \
    --cc=thomas.hellstrom@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: 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