Please consider these upstream CVE fixes for 6.6.y

Hi Greg and Sasha,

Thanks for carrying the earlier 6.1.y fixes into the newer trees. After dropping them, these four mainline fixes remain absent from 6.6.y and its current stable queue. Linux CNA records mark v6.6.155 affected for each CVE. Every newer tree has the first three fixes. The last remains absent from 6.12.y and 6.18.y.

I applied the current 6.6.y queue to v6.6.155, then cherry-picked this set without changes. The x86_64 allmodconfig kernel and modules built without warnings with WERROR enabled. The PowerPC target compiled cleanly after I enabled the related config options. I did not runtime-test these fixes.

Could you please queue these commits for 6.6.y?

0f022d32c3eca477fbf79a205243a6123ed0fe11 (CVE-2024-27010)
net/sched: Fix mirred deadlock on device recursion
d7f01649f4eaf1878472d3d3f480ae1e50d98f6c (CVE-2024-39478)
crypto: starfive - Do not free stack buffer
d48e1dea3931de64c26717adc2b89743c7ab6594 (CVE-2024-50225)
btrfs: fix error propagation of split bios
dbc30a57bd8e026995e9fa8e8c31cffd18542c01 (CVE-2026-64070)
powerpc/hv-gpci: fix preempt count leak in sysfs show paths

I used Codex to compare the histories and run the builds. I checked the final list against vulns.git, the branch history, and stable-queue.

Thanks,
Artem Dinaburg
Trail of Bits

Queued the mirred deadlock fix and its noop_qdisc owner follow-up plus the starfive fix for 6.6.y, and the hv-gpci fix for 6.18.y, 6.12.y and 6.6.y. Thanks.

Dropped the btrfs one - it doesn't build on 6.6 riscv, 1-byte cmpxchg isn't supported there.

Thanks,
Sasha

Queued the mirred deadlock fix and its noop_qdisc owner follow-up, the starfive fix, and the hv-gpci fix for 6.6.y (the hv-gpci fix also went to 6.18.y and 6.12.y). Dropped the btrfs one as it doesn't build on 6.6 riscv.

Thanks,
Sasha