cros-disks should mount iso9660 via FUSE |
|||||
Issue descriptionsplitting off from issue 787303 atm iso9660 is mounted using the kernel driver. that means people can download any file into Downloads and get it read/parsed directly by the kernel FS layer ... any bugs in that driver means immediate kernel access. libarchive has support for iso9660 and many common extensions: https://github.com/libarchive/libarchive/wiki/FormatISO9660 that means my WGU extension should be able to mount it via NaCl: https://chrome.google.com/webstore/detail/mljpablpddhocfbnokacjggdbmafjnon is there any reason to keep this enabled at the OS level ? metrics (in issue 787303) indicate this is lowish.
,
Nov 21 2017
gotcha, so WGU is already the way to mount an ISO image in the Files app. we still have the same attack surface, but it's a little bit harder (burn the malformed image, then plug it in via USB). replumbing the raw device node to chronos/Files app is probably not an improvement (due to ioctl exposure). so is the best we can do here is move to a fuse implementation like archivemount ? http://www.cybernoia.de/software/archivemount/ it uses libarchive to parse iso9660 images.
,
Feb 22 2018
,
Feb 28 2018
,
Jul 26
Is this issue still valid? We need iso9660 kernel support for physical CD disks (not images). I can't see a good way (I can see bad ways) or reason to remove it.
,
Jul 26
it's still valid in the sense that we're exposing the kernel directly to removable media which is bad for security. we shouldn't have a system where inserting a CD-ROM into a USB drive can result in immediate & full system exploit. i'm not saying there are any such known bugs today, but if there were, we have 0 layers of defense here.
,
Jul 27
We do this for FAT32, ext2/3/4, UDF, and HFS+. Is iso9660 any different from those?
,
Jul 27
every user that hits the kernel directly is a problem. we should have a bug per filesystem to migrate them as needed. I started with iso9660 as I had hoped to just delete it and be done. |
|||||
►
Sign in to add a comment |
|||||
Comment 1 by benchan@chromium.org
, Nov 21 2017