This series makes the AMD HSMP driver safe against concurrent probe/remove
of its per-socket devices and against the data plane (open /dev/hsmp fds and
hwmon/sysfs reads) racing socket teardown.
The ACPI front-end binds one platform device per socket but shares a single
socket array and a single /dev/hsmp misc device across them, while the data
plane issues mailbox messages with no coordination with driver teardown.
misc_deregister() does not drain already-open fds, so an in-flight message
can touch a freed socket array or an unmapped mailbox on unbind.
A single rw_semaphore, hsmp_sock_rwsem, serializes everything: the data
plane takes it for read so messages run concurrently with each other, and
probe and remove take it for write to bring sockets up and tear them down
while excluding and draining the data plane. The fix is built up in small,
bisectable steps:
1. Introduce hsmp_sock_rwsem and hold it for write across ACPI probe/remove
so concurrent per-socket probes cannot race the bring-up handshake or
the one-time socket-array allocation.
2. Map the metric table with ioremap() and release it via a devres action,
so its lifetime is no longer pinned to a single per-socket devres scope.
3. Serialize the per-socket metric-table fill-and-copy with a mutex.
4. Clear mdev.this_device on deregister (independent hygiene fix that the
next patch relies on to track /dev/hsmp registration).
5. Track shared socket ownership with a kref and a single coordinated
release helper, drop the is_probed flag and unparent /dev/hsmp on the
ACPI path.
6. Add the read side of hsmp_sock_rwsem to the data plane: split the send
into hsmp_send_message_locked() (asserts the rwsem is held) and
hsmp_send_message() (wraps it in guard(rwsem_read)). Route the probe-only
senders through the locked variant so probe, holding the write lock, does
not recurse on the rwsem.
Each patch builds on its own and the series is checkpatch --strict clean.
Changes since v6:
- Comment-only fixes addressing review feedback on patch 6: the code
comments no longer describe the data plane as "lock-free". After this
series hsmp_send_message() takes hsmp_sock_rwsem for read, so the data
plane is gated by a lock (the read side simply does not exclude other
readers); calling it lock-free contradicted the rwsem documentation and
was misleading. Reworded hsmp_send_message() (patch 6, hsmp.c) and the
kref-release / probe / remove teardown comments (patch 5, acpi.c), and
dropped the "lock-free" wording from the patch 1 and patch 6 changelogs.
No functional change.
Changes since v5:
- Track the shared ACPI sockets with a kref instead of a hand-rolled
unsigned int counter (patch 5), as suggested on v5. The first probe
kref_init()s it, each further probe kref_get()s and every remove or
probe failure kref_put()s; the last put runs hsmp_acpi_sock_release() as
the kref release callback. get/put still happen under hsmp_sock_rwsem
held for write, so the atomic is not strictly needed, but kref gives the
clearer interface.
- Keep the rwsem read side for the data plane rather than switching it to
kref_get_unless_zero(). The kref tracks the shared socket array, but each
socket's mailbox is a per-device devm_ioremap_uc() mapping that devres
unmaps on that socket's individual (non-final) unbind while the array and
the kref are still alive. An array-level kref would keep the array from
being freed but would not stop that per-socket mailbox unmap, so the read
side is what serializes the sock->dev gate and the MMIO access against a
concurrent per-socket teardown. A kref_get_unless_zero() data plane would
need a per-socket reference tied to each mailbox's lifetime, a larger
redesign; noted for a possible follow-up.
Muralidhara M K (6):
platform/x86/amd/hsmp: Serialize ACPI HSMP probe and remove with an
rwsem
platform/x86/amd/hsmp: Map the metric table with ioremap() and unmap
it explicitly
platform/x86/amd/hsmp: Serialize per-socket metric table reads with a
mutex
platform/x86/amd/hsmp: Clear mdev.this_device on deregister
platform/x86/amd/hsmp: ACPI HSMP refcounted sockets and coordinated
release
platform/x86/amd/hsmp: Serialize the data plane against socket
teardown
drivers/platform/x86/amd/hsmp/acpi.c | 146 +++++++++++++++++++++++++++++++----
drivers/platform/x86/amd/hsmp/hsmp.c | 112 +++++++++++++++++++++++++--
drivers/platform/x86/amd/hsmp/hsmp.h | 14 +++-
drivers/platform/x86/amd/hsmp/plat.c | 38 ++++++++-
4 files changed, 287 insertions(+), 23 deletions(-)
base-commit: ff7836fa850c2f815bc219f1e48f6ec8699f4ae7
--
2.34.1