Re: [Linux-kernel-mentees] [PATCH 15/16] vc_screen: extract vcs_read_buf_header

Re: [Linux-kernel-mentees] [PATCH 15/16] vc_screen: extract vcs_read_buf_header

[Peilin Ye]

Hi all,

Link: https://syzkaller.appspot.com/bug?id=f332576321998d36cd07d09c9c1268cfed1895c9

As reported by syzbot, vcs_read_buf() is overflowing `con_buf16`, since
this patch removed the following check:

-		if (count > CON_BUF_SIZE) {
-			count = CON_BUF_SIZE;
-			filled = count - pos;
-		}

Decreasing `count` by `min(HEADER_SIZE - pos, count)` bypasses this check.
Additionally, this patch also removed updates to `skip` and `filled`.

What should we do in order to fix it?

Thank you,
Peilin Ye
_______________________________________________
Linux-kernel-mentees mailing list
Linux-kernel-mentees@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees

Re: [Linux-kernel-mentees] [PATCH 15/16] vc_screen: extract vcs_read_buf_header

[Greg KH]

On Tue, Aug 25, 2020 at 12:48:04PM -0400, Peilin Ye wrote:
> Hi all,
> 
> Link: https://syzkaller.appspot.com/bug?id=f332576321998d36cd07d09c9c1268cfed1895c9
> 
> As reported by syzbot, vcs_read_buf() is overflowing `con_buf16`, since
> this patch removed the following check:
> 
> -		if (count > CON_BUF_SIZE) {
> -			count = CON_BUF_SIZE;
> -			filled = count - pos;
> -		}
> 
> Decreasing `count` by `min(HEADER_SIZE - pos, count)` bypasses this check.
> Additionally, this patch also removed updates to `skip` and `filled`.
> 
> What should we do in order to fix it?

This patch is already reverted, and it has been discussed a bit as to
how to do this properly if you look at the email where this was reported
to us.

thanks,

greg k-h
_______________________________________________
Linux-kernel-mentees mailing list
Linux-kernel-mentees@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees

Re: [Linux-kernel-mentees] [PATCH 15/16] vc_screen: extract vcs_read_buf_header

[Peilin Ye]

On Tue, Aug 25, 2020 at 06:59:35PM +0200, Greg KH wrote:
> On Tue, Aug 25, 2020 at 12:48:04PM -0400, Peilin Ye wrote:
> > Hi all,
> > 
> > Link: https://syzkaller.appspot.com/bug?id=f332576321998d36cd07d09c9c1268cfed1895c9
> > 
> > As reported by syzbot, vcs_read_buf() is overflowing `con_buf16`, since
> > this patch removed the following check:
> > 
> > -		if (count > CON_BUF_SIZE) {
> > -			count = CON_BUF_SIZE;
> > -			filled = count - pos;
> > -		}
> > 
> > Decreasing `count` by `min(HEADER_SIZE - pos, count)` bypasses this check.
> > Additionally, this patch also removed updates to `skip` and `filled`.
> > 
> > What should we do in order to fix it?
> 
> This patch is already reverted, and it has been discussed a bit as to
> how to do this properly if you look at the email where this was reported
> to us.

Ah, syzbot has reported this issue as two different bugs...I started
looking into it without noticing your discussion under another thread.

Sorry,
Peilin Ye
_______________________________________________
Linux-kernel-mentees mailing list
Linux-kernel-mentees@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees