linux-kernel - Re: Is: virtio_gpu_object_shmem_init issues? Was:Re: upstream boot error: general protection fault in swiotlb

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: <20200824151300.GA28575@lst.de>
Date:   Mon, 24 Aug 2020 17:13:00 +0200
From:   Christoph Hellwig <hch@....de>
To:     Konrad Rzeszutek Wilk <konrad.wilk@...cle.com>
Cc:     syzbot <syzbot+3f86afd0b1e4bf1cb64c@...kaller.appspotmail.com>,
        thomas.tai@...cle.com, airlied@...hat.com, kraxel@...hat.com,
        virtualization@...ts.linux-foundation.org,
        dri-devel@...ts.freedesktop.org, hch@....de,
        iommu@...ts.linux-foundation.org, linux-kernel@...r.kernel.org,
        m.szyprowski@...sung.com, robin.murphy@....com,
        syzkaller-bugs@...glegroups.com
Subject: Re: Is: virtio_gpu_object_shmem_init  issues? Was:Re: upstream
 boot error: general protection fault in swiotlb_map

On Mon, Aug 24, 2020 at 11:06:51AM -0400, Konrad Rzeszutek Wilk wrote:
> So it fails at
> 
> 683                 dev_WARN_ONCE(dev, 1,                                           
> 684                         "swiotlb addr %pad+%zu overflow (mask %llx, bus limit %llx).\n",
> 685                         &dma_addr, size, *dev->dma_mask, dev->bus_dma_limit);   
> 
> 
> which makes no sense to me as `dev` surely exists. I can see in the console log:
> 
> virtio-pci 0000:00:01.0: vgaarb: deactivate vga console
> 
> So what gives?

Well, look at the if around the WARN_ON - dma_capable failed on the
swiotlb buffer. This means the virtio drm thingy has a dma mask
(either the actual one set by the driver, or bus_dma_mask), which isn't
enough to address the swiotlb buffer.