On Thu, 12 Nov 2020 at 17:15, Greg KH gregkh@linuxfoundation.org wrote:
On Thu, Nov 12, 2020 at 10:12:01AM +1100, Anand K. Mistry wrote:
On Wed, 11 Nov 2020 at 23:23, Greg KH gregkh@linuxfoundation.org wrote:
On Wed, Nov 11, 2020 at 11:09:13PM +1100, Anand K. Mistry wrote:
Hi,
I'm requesting a stable merge for commit 1978b3a53a74e3230cd46932b149c6e62e832e9a ("x86/speculation: Allow IBPB to be conditionally enabled on CPUs with always-on STIBP") into the stable branch for 5.4. Note, the commit is already queued for inclusion into the next 5.9 stable release (https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree...).
The patch fixes an issue where a Spectre-v2-user mitigation could not be enabled via prctl() on certain AMD CPUs. The issue was introduced in commit 21998a351512eba4ed5969006f0c55882d995ada ("x86/speculation: Avoid force-disabling IBPB based on STIBP and enhanced IBRS.") which was merged into the 5.4 stable branch as commit 6d60d5462a91eb46fb88b016508edfa8ee0bc7c8. This commit also exists in 4.19, 4.14, 4.9, and 4.4, so those kernels are also likely affected by this bug.
As I asked when I sent out a "FAILED:" message for this patch, if someone wants it backported to older kernels, they will need to provide the backported versions of it, as the patch does not apply cleanly as-is.
Can you please do that?
Oh, I didn't get that message. I'll prepare a backport.
You didn't have to get the email, but I would assume that you at least tested the backport if you asked for it to happen, right? :)
The conflict was so trivial (a single newline in a comment) I didn't really think about it. And yes, the patch is well tested against 5.4 (which is my target kernel), and 4.14.