drop SECURITY_FILE_CAPABILITIES? (fwd)

[James Morris]

How might this affect the Fedora kernel?

---------- Forwarded message ----------
Date: Tue, 10 Nov 2009 08:07:39 -0600
From: Serge E. Hallyn <serue at us.ibm.com>
To: lkml <linux-kernel at vger.kernel.org>
Cc: linux-security-module at vger.kernel.org, Andrew Morgan <morgan at kernel.org>,
    Steve Grubb <sgrubb at redhat.com>, Kees Cook <kees.cook at canonical.com>,
    Andreas Gruenbacher <agruen at suse.de>,
    Michael Kerrisk <mtk.manpages at gmail.com>,
    George Wilson <gcwilson at us.ibm.com>
Subject: drop SECURITY_FILE_CAPABILITIES?

Hey,

Just a probe to see what people think.  I've seen two cases
in about the last month where software was confounded by
an assumption that prctl(PR_CAPBSET_DROP, CAP_SOMETHING)
would succeed if privileged, but not handling the fact
that SECURITY_FILE_CAPABILITIES=n means you can't do that.

Are we at the point yet where we feel we can get rid of
the SECURITY_FILE_CAPABILITIES=n case?

Note that there is a boot arg no_file_caps which prevents
file capabilities from being used if SECURITY_FILE_CAPABILITIES=y.
I think that's the case most users will care about, whereas the
remaining differences between CONFIG_SECURITY_FILE_CAPABILITIES=y
and =n are that with CONFIG_SECURITY_FILE_CAPABILITIES=y :

	(1) certain security hooks (task_setscheduler, task_setioprio, and
	task_setnice) do capability set comparisions,
	(2) it is possible to drop capabilities from the bounding set,
	(3) it is possible to set per-task securelevels,
	(4) and it is possible to add any capability to your inheritable
	set if you have CAP_SETPCAP.

Does anyone know of cases where CONFIG_SECURITY_FILE_CAPABILITIES=n
is still perceived as useful?

thanks,
-serge
--
To unsubscribe from this list: send the line "unsubscribe linux-security-module" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html