linux-kernel - Re: [PATCH] drm/hyperv: Don't rely on screen_info.lfb

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]

Date:   Sun, 11 Sep 2022 09:21:19 -0700
From:   Saurabh Singh Sengar <ssengar@...ux.microsoft.com>
To:     Thomas Zimmermann <tzimmermann@...e.de>
Cc:     ssengar@...rosoft.com, drawat.floss@...il.com, airlied@...ux.ie,
        daniel@...ll.ch, linux-hyperv@...r.kernel.org,
        dri-devel@...ts.freedesktop.org, linux-kernel@...r.kernel.org,
        mikelley@...rosoft.com
Subject: Re: [PATCH] drm/hyperv: Don't rely on screen_info.lfb_base for Gen1
 VMs

On Sat, Sep 10, 2022 at 08:11:24PM +0200, Thomas Zimmermann wrote:
> Hi
> 
> Am 09.09.22 um 16:43 schrieb Saurabh Sengar:
> >hyperv_setup_vram tries to remove conflicting framebuffer based on
> >'screen_info'. As observed in past due to some bug or wrong setting
> >in grub, the 'screen_info' fields may not be set for Gen1, and in such
> >cases drm_aperture_remove_conflicting_framebuffers will not do anything
> >useful.
> >For Gen1 VMs, it should always be possible to get framebuffer
> >conflict removed using PCI device instead.
> >
> >Fixes: a0ab5abced55 ("drm/hyperv : Removing the restruction of VRAM allocation with PCI bar size")
> >Signed-off-by: Saurabh Sengar <ssengar@...ux.microsoft.com>
> >---
> >  drivers/gpu/drm/hyperv/hyperv_drm_drv.c | 24 ++++++++++++++++++++----
> >  1 file changed, 20 insertions(+), 4 deletions(-)
> >
> >diff --git a/drivers/gpu/drm/hyperv/hyperv_drm_drv.c b/drivers/gpu/drm/hyperv/hyperv_drm_drv.c
> >index 6d11e7938c83..b0cc974efa45 100644
> >--- a/drivers/gpu/drm/hyperv/hyperv_drm_drv.c
> >+++ b/drivers/gpu/drm/hyperv/hyperv_drm_drv.c
> >@@ -73,12 +73,28 @@ static int hyperv_setup_vram(struct hyperv_drm_device *hv,
> >  			     struct hv_device *hdev)
> >  {
> >  	struct drm_device *dev = &hv->dev;
> >+	struct pci_dev *pdev;
> >  	int ret;
> >-	drm_aperture_remove_conflicting_framebuffers(screen_info.lfb_base,
> >-						     screen_info.lfb_size,
> >-						     false,
> >-						     &hyperv_driver);
> >+	if (efi_enabled(EFI_BOOT)) {
> >+		drm_aperture_remove_conflicting_framebuffers(screen_info.lfb_base,
> >+							     screen_info.lfb_size,
> 
> Using screen_info here seems wrong in any case. You want to remove
> the framebuffer devices that conflict with your driver, which might
> be unrelated to screen_info. AFAICT the correct solution would
> always retrieve the PCI device for removal (i.e., always do the else
> branch).

In a Gen2 VM, the Hyper-V frame buffer device is presented only as a VMbus device.
It's not presented as a PCI device like it is in a Gen1 VM. This would have worked
if we had the frame buffer device available as PCI device in Gen2 but unfortunately
thats not the case here.

Regards,
Saurabh

> 
> Best regard
> Thomas
> 
> >+							     false,
> >+							     &hyperv_driver);
> >+	} else {
> >+		pdev = pci_get_device(PCI_VENDOR_ID_MICROSOFT, PCI_DEVICE_ID_HYPERV_VIDEO, NULL);
> >+		if (!pdev) {
> >+			drm_err(dev, "Unable to find PCI Hyper-V video\n");
> >+			return -ENODEV;
> >+		}
> >+
> >+		ret = drm_aperture_remove_conflicting_pci_framebuffers(pdev, &hyperv_driver);
> >+		pci_dev_put(pdev);
> >+		if (ret) {
> >+			drm_err(dev, "Not able to remove boot fb\n");
> >+			return ret;
> >+		}
> >+	}
> >  	hv->fb_size = (unsigned long)hv->mmio_megabytes * 1024 * 1024;
> 
> -- 
> Thomas Zimmermann
> Graphics Driver Developer
> SUSE Software Solutions Germany GmbH
> Maxfeldstr. 5, 90409 Nürnberg, Germany
> (HRB 36809, AG Nürnberg)
> Geschäftsführer: Ivo Totev