#1120972 Linux side mitigations for CVE-2025-62626

Package:
src:linux
Source:
src:linux
Submitter:
Mario Limonciello
Date:
2025-11-19 06:18:01 UTC
Severity:
normal
Tags:
#1120972#5
Date:
2025-11-18 22:01:09 UTC
From:
To:
Dear Maintainer,

CVE-2025-62626 (AKA AMD-SB-7055) is a vulnerability in the instruction RDSEED's 16-bit and 32-bit returns. It affects AMD Zen 5 hardware.
As there are coordinations needed from both linux-firmware and linux I wanted to provide a comprehensive overview of everything.
[Some of these kernel patches are already landed in Debian - just want you to have the whole picture.]

https://www.amd.com/en/resources/product-security/bulletin/amd-sb-7055.html

There are multiple things that need to be done about this vulnerability.

1) Updated linux-firmware microcode has been upstreamed for Zen5 hardware. This affects both client and datacenter hardware.

https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/commit/?id=6167e5566900cf236f7a69704e8f4c441bc7212a

2) A mitigation has been put in place in the kernel for when there is NOT an updated microcode available. This disables the advertisement of the RDSEED instruction to userspace and prevents it's use in the kernel. As there is no feature flag for 16, 32 and 64 it unfortunately disables all of them.

https://git.kernel.org/torvalds/c/607b9fb2ce248
https://git.kernel.org/torvalds/c/f1fdffe0afea0

3) Additional models need to be added to entry sign checking. In order to apply the fix for rdseed the base information for entry sign must be present.

https://git.kernel.org/torvalds/c/8a9fb5129e8e6
https://git.kernel.org/torvalds/c/d23550efc6800
https://git.kernel.org/torvalds/c/dd14022a7ce96

4) Allow client systems to use RDSEED.

https://git.kernel.org/torvalds/c/e1a97a627cd01

#1120972#12
Date:
2025-11-19 05:47:18 UTC
From:
To:
Control: found -1 6.12.57-1
Control: found -1 6.17.7-1
Control: fixed -1 6.18~rc6-1~exp1
Hi Mario,

Thank you indeed yes we are aware (and usually I would ask to not fill
explicit CVE bugs for src:linux, we have a separate tracking for that,
but here the CVE is more associated with amd64-microcode, cf.
#1120005, and mitigations exists in the Linux kernel, this my
understanding so I have slightly redacted the subject).

We follow stable series of usptream Linux in Debian so as long it is
guaranteed fixes land in the needed stable series we are done.

Here a summary though were we stand:

This is to be tracked in #1120005.

In 6.18-rc4 (and backported to 6.12.58 and 6.17.8).

In 6.18-rc5 (and already backported to 6.17.8).

In 6.18-rc4 (not yet backported/released in stable series).

In 6.18-rc5 (already released as well in 6.12.58 and 6.17.8)

In 6.18-rc6 (but not yet in released stable series)

In 6.18-rc6 (but not yet in released stable series).

I assume hwere needed you are making sure Greg or Sasha are picking up
the needed changes for sthe stable series?

Regards,
Salvatore

#1120972#27
Date:
2025-11-19 05:50:46 UTC
From:
To:
close 1120972
# make sure bug only get archived when fixed as well in trixie and forky
tags 1120972 + trixie forky sid
thanks