|
Message-id: <fae0de0ca6ac.54f3ef00@langara.bc.ca> Date: Mon, 02 Mar 2015 05:02:56 +0000 (GMT) From: Steven Stewart-Gallus <sstewartgallus00@...angara.bc.ca> To: Rich Felker <dalias@...c.org> Cc: oss-security@...ts.openwall.com, ryao@...too.org Subject: Re: CVE request: Linux kernel silently ignores MS_RDONLY for bind mounts Hello, I suppose it's time I gave my opinion on this matter. Personally, I am ambivalent about whether this really deserves a CVE (or if the CVE should be with the Linux kernel or with the applications that misuse this API) as I feel it is the responsibility of API users like LXC and systemd to make sure that they aren't misusing these interfaces but I would still like this feature to be implemented and I will explain why. For my own needs (with my own project at https://gitorious.org/linted/linted) I sandbox processes without raising privileges by means such as setuid applications and so can only map uids and gids to the current user. However, I still need to prevent certain processes from writing to the user's home directory and as such need to mount the /home hierarchy read only and recursively. Mostly though this is not a big problem for me because I only need to mount the user's home directory when developing (because I need to run binaries that are built inside the user's home directory). Also, there is the possibility of bind mounting special hierarchies such as /dev, /proc and /sys read only (these are not just one filesystem but need to be bound recursively) but I don't consider this a strong use case. Thank you, Steven Stewart-Gallus
Powered by blists - more mailing lists
Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.