We're currently using Linux 3.0.12 with Cristoph's
xfs-bulletproof-sync patch and it seems to be working very well for
us. Unfortunately this kernel is vulnerable to the recent
CVE-2012-0056 no permission checking on writes to /proc/(pid)/mem
local root exploit, so we've got to leave it behind.
I see that the newest recommended stable kernel on kernel.org is
3.2.9. Have there been any major changes to XFS between 3.0.12 and
3.2.9 that would be considered "risky" in a production environment? Is
there any reason why we shouldn't be using 3.2.9 in production? We got
bit pretty hard by various sync failure bugs in 2.6.38 that resulted
in major data loss - so I want to make absolutely sure there aren't
any "snakes in the grass" before we try a newer kernel.
If 3.2.9 is not suitable, is there at least a kernel in 3.2-train that
is fairly safe as far as XFS goes?
I assume the xfs-bulletproof-sync patch has already been committed to
the code base in 3.2-train, so we shouldn't have to worry about that
I'm not subscribed to the list; please CC any replies to my personal
e-mail and expect some delay in response; I'm just watching the
mailing list through the archives.