On Fri, Oct 16, 2020 at 12:01 PM Daniel Burgener dburgener@linux.microsoft.com wrote:
On 10/16/20 11:49 AM, Sasha Levin wrote:
On Fri, Oct 16, 2020 at 05:44:43PM +0200, Greg KH wrote:
On Fri, Oct 16, 2020 at 11:38:23AM -0400, Sasha Levin wrote:
On Fri, Oct 16, 2020 at 05:01:20PM +0200, Greg KH wrote:
On Fri, Oct 16, 2020 at 09:48:31AM -0400, Daniel Burgener wrote:
v2: Include all commits from original series, and include commit
ids
This is a backport for stable of my series to fix a race
condition in
selinuxfs during policy load:
Has this race condition always been present, or is this a regression that is being fixed from previously working kernels?
So this issue has always been there, but:
If it's always been present, why not just use 5.9 to solve it?
Because it was merged for 5.10 rather than 5.9, which is a few months out, so Daniel is looking to see if we can have it in 5.8/5.4 to close the gap.
I would have to wait for 5.10-rc1 at the earliest, and get the selinux maintainers ack for this :)
No objections; if the selinux folks feel unhappy with this it'll just wait for 5.10.
Yes, that's fine from my end as well. We can carry this series out of tree in the interim. Thanks!
I tend to be pretty conservative when it comes to backporting patches to -stable, and since this is both a) big and b) fixes a problem that has existed since the dawn of selinuxfs (and possibly longer <g>) I think the smart thing to do is to wait for v5.10.