[PATCH v3 0/9] pci: fix UAF and TOCTOU related to dynamic ID

[Gary Guo <gary@garyguo.net>]

While working on improving the Rust abstractions [1], Sashiko reported that
an existing UAF issue related to dynamic ID, which I find to be genuine.
When taking a look at the code I also find a TOCTOU issue where the
existence check of dynamic ID happens in a separate critical section as the
actual insertion. This series fix both issues.

There are two exported functions "pci_match_id" and "pci_add_dynid" which I
have to tweak to implement this cleanly; I created separate "do_xxx"
functions to keep the existing APIs because they all have multiple users.

There're a few existing users which stores their pci_device_id argument in
probe callback. This is a bad pattern because nothing except driver_data
inside pci_device_id is what they want; actual ID information can be
retrieved from pci_dev instead.

There are two users that performs pointer arithmetic on the pci_device_id;
these are also problematic with dynamic ID and driver_override, so fix them
as well.

I've used the following coccinelle script to flag all cases where the
pci_device_id is used other than reading its fields.

@usage@
identifier fn, id;
position p;
@@
  fn(..., struct pci_device_id *id, ...)
  {
    ...
    id@p
    ...
  }

// Due to cocci isomorphism this needs to be explicit
@bad@
identifier fn, id;
type T;
position usage.p;
@@
  fn(..., struct pci_device_id *id, ...)
  {
    ...
    (T*)id@p
    ...
  }

// Good use cases
@good@
identifier fn, id, fld;
expression E;
position usage.p;
@@
  fn(..., struct pci_device_id *id, ...)
  {
    ...
(
    id@p->fld
|
    E(..., id@p, ...)
|
// Redundant checks, but ignore
    !id@p
|
// Redundant checks, but ignore
    id ? ... : ...
)
    ...
  }

@script:python depends on usage && (bad || !good)@
p << usage.p;
@@
coccilib.report.print_report(p[0], "suspicious use of pci_device_id")

Link: https://lore.kernel.org/all/20260618-id_info-v1-0-96af1e559ef9@garyguo.net/ [1]
Link: https://lore.kernel.org/all/20260619170503.518F61F00A3A@smtp.kernel.org/ [2]

---
Changes in v3:
- Fix users which uses pci_device_id for pointer arithmetic. (Sashiko)
- Convert to scoped_guard. (Danilo)
- For static IDs, still give out static pointers and avoid making a copy.
- Link to v2: https://patch.msgid.link/20260630-pci_id_fix-v2-0-b834a98c0af2@garyguo.net

Changes in v2:
- Fix users which store pci_device_id.
- Clarify in probe documentation about the lifetime of pci_device_id
  parameter.
- Dynamic ID conflict check now ignores override_only. (Sashiko)
- Link to v1: https://patch.msgid.link/20260626-pci_id_fix-v1-0-a35c803f1b95@garyguo.net

---
Gary Guo (9):
      ata: don't store pci_device_id
      nsp32: don't store pci_device_id
      ipack: tpci200: don't store pci_device_id
      mlxsw: don't store pci_device_id
      agp/via: don't rely on address of pci_device_id
      agp/amd-k7: don't rely on address of pci_device_id
      pci: make pci_match_one_device match on ID instead of device
      pci: fix dyn_id add TOCTOU
      pci: fix UAF when probe runs concurrent to dyn ID removal

 drivers/ata/ata_generic.c                 |   6 +-
 drivers/char/agp/amd-k7-agp.c             |  26 +--
 drivers/char/agp/via-agp.c                | 308 +++++++-----------------------
 drivers/ipack/carriers/tpci200.c          |   1 -
 drivers/ipack/carriers/tpci200.h          |   1 -
 drivers/net/ethernet/mellanox/mlxsw/pci.c |  11 +-
 drivers/pci/pci-driver.c                  | 193 ++++++++++---------
 drivers/pci/pci.h                         |  36 +++-
 drivers/pci/search.c                      |   6 +-
 drivers/scsi/nsp32.c                      |   8 +-
 drivers/scsi/nsp32.h                      |   8 +-
 include/linux/pci.h                       |   1 +
 12 files changed, 230 insertions(+), 375 deletions(-)
---
base-commit: 2b763db0c2763d6bf73d7d3e69665222d1f377cf
change-id: 20260626-pci_id_fix-83eaec007674

Best regards,
--  
Gary Guo <gary@garyguo.net>