linux-kernel - RE: [PATCH 1/1] fbdev: hyperv_fb: iounmap() the correct memory when removing a device

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:
 <SN6PR02MB4157C1DF0A0101EEF4CA79E2D4FF2@SN6PR02MB4157.namprd02.prod.outlook.com>
Date: Thu, 13 Feb 2025 01:35:22 +0000
From: Michael Kelley <mhklinux@...look.com>
To: Saurabh Singh Sengar <ssengar@...ux.microsoft.com>
CC: "haiyangz@...rosoft.com" <haiyangz@...rosoft.com>, "wei.liu@...nel.org"
	<wei.liu@...nel.org>, "decui@...rosoft.com" <decui@...rosoft.com>,
	"deller@....de" <deller@....de>, "weh@...rosoft.com" <weh@...rosoft.com>,
	"dri-devel@...ts.freedesktop.org" <dri-devel@...ts.freedesktop.org>,
	"linux-fbdev@...r.kernel.org" <linux-fbdev@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"linux-hyperv@...r.kernel.org" <linux-hyperv@...r.kernel.org>
Subject: RE: [PATCH 1/1] fbdev: hyperv_fb: iounmap() the correct memory when
 removing a device

From: Saurabh Singh Sengar <ssengar@...ux.microsoft.com> Sent: Monday, February 10, 2025 8:52 AM
> 
[snip]
> > > >
> > > > While we are at it, I want to mention that I also observed below WARN
> > > > while removing the hyperv_fb, but that needs a separate fix.
> > > >
> > > >
> > > > [   44.111220] WARNING: CPU: 35 PID: 1882 at drivers/video/fbdev/core/fb_info.c:70 framebuffer_release+0x2c/0x40
> > > > < snip >
> > > > [   44.111289] Call Trace:
> > > > [   44.111290]  <TASK>
> > > > [   44.111291]  ? show_regs+0x6c/0x80
> > > > [   44.111295]  ? __warn+0x8d/0x150
> > > > [   44.111298]  ? framebuffer_release+0x2c/0x40
> > > > [   44.111300]  ? report_bug+0x182/0x1b0
> > > > [   44.111303]  ? handle_bug+0x6e/0xb0
> > > > [   44.111306]  ? exc_invalid_op+0x18/0x80
> > > > [   44.111308]  ? asm_exc_invalid_op+0x1b/0x20
> > > > [   44.111311]  ? framebuffer_release+0x2c/0x40
> > > > [   44.111313]  ? hvfb_remove+0x86/0xa0 [hyperv_fb]
> > > > [   44.111315]  vmbus_remove+0x24/0x40 [hv_vmbus]
> > > > [   44.111323]  device_remove+0x40/0x80
> > > > [   44.111325]  device_release_driver_internal+0x20b/0x270
> > > > [   44.111327]  ? bus_find_device+0xb3/0xf0
> > > >
> > >
> > > Thanks for pointing this out. Interestingly, I'm not seeing this WARN
> > > in my experiments. What base kernel are you testing with? Are you
> > > testing on a local VM or in Azure? What exactly are you doing
> > > to create the problem? I've been doing unbind of the driver,
> > > but maybe you are doing something different.
> > >
> > > FWIW, there is yet another issue where after doing two unbind/bind
> > > cycles of the hyperv_fb driver, there's an error about freeing a
> > > non-existent resource. I know what that problem is, and it's in
> > > vmbus_drv.c. I'll be submitting a patch for that as soon as I figure out
> > > a clean fix.
> > >
> > > Michael
> >
> > This is on local Hyper-V. Kernel: 6.14.0-rc1-next-20250205+
> > I run below command to reproduce the above error:
> > echo "5620e0c7-8062-4dce-aeb7-520c7ef76171" >
> /sys/bus/vmbus/devices/5620e0c7-8062-4dce-aeb7-520c7ef76171/driver/unbind
> >
> > When hvfb_remove is called I can see the refcount for framebuffer is 2 when ,
> > I expect it to be 1. After unregistering this framebuffer there is still 1 refcount
> > remains, which is the reason for this WARN at the time of framebuffer_release.
> >
> > I wonder who is registering/using this extra framebuffer. Its not hyperv_drm or
> > hyperv_fb IIUC.
> >
> > - Saurabh
> 
> Here are more details about this WARN:
> 
> Xorg opens `/dev/fb0`, which increases the framebuffer's reference
> count, as mentioned above.  As a result, when unbinding the driver,
> this WARN is expected, indicating that the framebuffer is still in use.
> 
> I am open to suggestion what could be the correct behavior in this case.
> There acan be two possible options:
> 
>  1. Check the framebuffer reference count and prevent the driver from
>     unbinding/removal.
> OR
> 
>  2. Allow the driver to unbind while issuing this WARN. (Current scenario)
> 

>From looking at things and doing an experiment, I think there's a 3rd
option, which gets rid of the of the WARN while still allowing the unbind.

The experiment is to boot Linux in a Gen2 Hyper-V guest with both the
Hyper-V FB and Hyper-V DRM modules removed. In this case, the
generic EFI framebuffer driver (efifb) should get used. With this driver,
a program can open /dev/fb0, and while it is open, unbind the efifb
driver (which is in /sys/bus/platform/drivers/efi-framebuffer).
Interestingly, there's no WARN generated. But when the hyperv_fb
driver is loaded and used, the WARN *is* generated, as you observed.

So I looked at the code for efifb.  It does the framebuffer_release()
call in a function that hyperv_fb doesn't have. Based on the comments
in efifb.c, we need a similar function to handle the call to
framebuffer_release().  And the efifb driver also does the iounmap()
in that same function, which makes we wonder if the hyperv_fb
driver should do similarly. It will need a little more analysis to
figure that out.

You found the bug.  Do you want to work on fixing the hyperv_fb
driver? And maybe the Hyper-V DRM driver needs the same fix.
I haven't looked. Alternatively, if you are busy, I can work on the fix.
Let me know your preference.

Michael