]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commit
KVM: s390: fix race in gmap_make_secure()
authorClaudio Imbrenda <imbrenda@linux.ibm.com>
Fri, 28 Apr 2023 09:27:53 +0000 (11:27 +0200)
committerClaudio Imbrenda <imbrenda@linux.ibm.com>
Thu, 4 May 2023 16:26:27 +0000 (18:26 +0200)
commitc148dc8e2fa403be501612ee409db866eeed35c0
tree7f2c58e528c596849603797177ba515c69bbacf3
parent292a7d6fca33df70ca4b8e9b0d0e74adf87582dc
KVM: s390: fix race in gmap_make_secure()

Fix a potential race in gmap_make_secure() and remove the last user of
follow_page() without FOLL_GET.

The old code is locking something it doesn't have a reference to, and
as explained by Jason and David in this discussion:
https://lore.kernel.org/linux-mm/Y9J4P%2FRNvY1Ztn0Q@nvidia.com/
it can lead to all kind of bad things, including the page getting
unmapped (MADV_DONTNEED), freed, reallocated as a larger folio and the
unlock_page() would target the wrong bit.
There is also another race with the FOLL_WRITE, which could race
between the follow_page() and the get_locked_pte().

The main point is to remove the last use of follow_page() without
FOLL_GET or FOLL_PIN, removing the races can be considered a nice
bonus.

Link: https://lore.kernel.org/linux-mm/Y9J4P%2FRNvY1Ztn0Q@nvidia.com/
Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Fixes: 214d9bbcd3a6 ("s390/mm: provide memory management functions for protected KVM guests")
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Message-Id: <20230428092753.27913-2-imbrenda@linux.ibm.com>
arch/s390/kernel/uv.c