Skip to contentSkip to conversation

driver core: Disable driver overriding by default

PATCH In review Uwe Kleine-König 8 participants 20 messages Nov 20 – Nov 21

Subsystems (38)
lore
Share & tools
lore
#0 lore raw

v2 cover driver core: Disable driver overriding by default

Hello,

here comes v2 which addresses the review feedback I got for (implicit)
v1. See https://lore.kernel.org/linux-kernel/[email protected]
for the discussion about that previous submission.

The objective is to deny force binding a device to a driver that isn't prepared for that. There are many drivers that run in a null pointer exception when this happens and only very few drivers are supposed to be used like that. So the first patch implements an explicit opt-in for drivers and denies other matches; the remaining patches implement the opt-in for drivers that rely on it. This makes it unnecessary to fix individual drivers to explicitly refuse such bindings. See commit 1e50a0349012 ("iio: adc: mcp3564: reject devices without match data") for an example of such a driver-specific fix.

Changes since (implicit) v1:

- Rework parameter parsing to just use allow_driver_override on the kernel commandline
- Add review tags
- Add patches by Alex Williamson setting .support_driver_override in pci and vfio drivers
- Add patches for uio, spidev and a few more drivers setting .support_driver_override
- align wording in Documentation/admin-guide/tainted-kernels.rst with tools/debugging/kernel-chktaint

It is sensible to merge the driver core change (patch #1) together with the individual driver patches setting .support_driver_override because with the driver core change the driver override doesn't work any more for the unmodified drivers (unless the new kernel parameter is provided and a taint is accepted). So I suggest (once there is agreement that the patches are fine) to merge them all in one go and provide a stable branch in case a driver was missed to be adapted. I can create such a branch and ask Greg or Linus to pull it, or Greg can do that.

So if you're the maintainer of one of the drivers touched by patches #2 and later and assuming your ok with that patch, please provide your Ack (plus the required coordination if applicable) for not merging via your usual tree.

Best regards
Uwe

Alex Williamson (3):
PCI: Allow untainted driver_override for PCI stub drivers
vfio: Allow untainted use of driver_override s390/vfio: Allow untainted use of driver_override

Uwe Kleine-König (7):
driver core: Disable driver overriding by default
uio: Allow untainted use of driver_override
spi: spidev: Allow untainted use of driver_override
clk: imx: scu: Allow untainted use of driver_override
remoteproc: Allow untainted use of driver_override
slimbus: qcom-ngd-ctrl: Allow untainted use of driver_override
ASoC: samsung: i2s: Allow untainted use of driver_override

 .../admin-guide/kernel-parameters.txt         |  6 +++
 Documentation/admin-guide/tainted-kernels.rst | 54 ++++++++++---------
 drivers/base/bus.c                            | 51 ++++++++++++++++++
 drivers/clk/imx/clk-scu.c                     |  1 +
 drivers/pci/pci-pf-stub.c                     |  1 +
 drivers/pci/pci-stub.c                        |  1 +
 drivers/rpmsg/rpmsg_ctrl.c                    |  1 +
 drivers/rpmsg/rpmsg_ns.c                      |  5 +-
 drivers/s390/cio/vfio_ccw_drv.c               |  1 +
 drivers/s390/crypto/vfio_ap_drv.c             |  1 +
 drivers/slimbus/qcom-ngd-ctrl.c               |  1 +
 drivers/spi/spidev.c                          |  1 +
 drivers/uio/uio_dmem_genirq.c                 |  1 +
 drivers/uio/uio_hv_generic.c                  |  1 +
 drivers/uio/uio_pci_generic.c                 |  1 +
 drivers/uio/uio_pdrv_genirq.c                 |  1 +
 drivers/vfio/cdx/main.c                       |  1 +
 drivers/vfio/fsl-mc/vfio_fsl_mc.c             |  1 +
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c    |  3 +-
 drivers/vfio/pci/ism/main.c                   |  3 +-
 drivers/vfio/pci/mlx5/main.c                  |  3 +-
 drivers/vfio/pci/nvgrace-gpu/main.c           |  3 +-
 drivers/vfio/pci/pds/pci_drv.c                |  3 +-
 drivers/vfio/pci/qat/main.c                   |  3 +-
 drivers/vfio/pci/vfio_pci.c                   |  3 +-
 drivers/vfio/pci/virtio/main.c                |  3 +-
 drivers/vfio/pci/xe/main.c                    |  3 +-
 drivers/vfio/platform/vfio_amba.c             |  1 +
 drivers/vfio/platform/vfio_platform.c         |  1 +
 include/linux/device.h                        | 19 +------
 include/linux/device/driver.h                 |  2 +
 include/linux/panic.h                         |  3 +-
 include/trace/events/module.h                 |  3 +-
 kernel/panic.c                                |  3 +-
 sound/soc/samsung/i2s.c                       |  1 +
 tools/debugging/kernel-chktaint               |  9 ++++
 36 files changed, 144 insertions(+), 55 deletions(-)

base-commit: 7cfbf79c66c63e807ea452abf0546e6f6d6833a6

Signature · 1 line
2.56.0.rc1
#1 lore raw

PATCH v2 1/10 driver core: Disable driver overriding by default

Driver overriding is useful only in a very limited set of situations and then only with very few drivers that are designed to support that.

Disallow matching via driver_override unless the driver explicitly allows it or the safe guard is disabled using the "allow_driver_override=1" kernel parameter.

There are many drivers that don't expect to be force-bound and often doing it anyhow results in a NULL pointer exception. To make debugging such cases easier, add a new taint if such a forced match happens.

Implementation detail: device_match_driver_override() isn't a static inline any more. It grew a certain complexity (e.g. a pr_info()) and so was made a regular exported function.

---
 .../admin-guide/kernel-parameters.txt         |  6 +++
 Documentation/admin-guide/tainted-kernels.rst | 54 ++++++++++---------
 drivers/base/bus.c                            | 51 ++++++++++++++++++
 include/linux/device.h                        | 19 +------
 include/linux/device/driver.h                 |  2 +
 include/linux/panic.h                         |  3 +-
 include/trace/events/module.h                 |  3 +-
 kernel/panic.c                                |  3 +-
 tools/debugging/kernel-chktaint               |  9 ++++
 9 files changed, 105 insertions(+), 45 deletions(-)
Documentation/admin-guide/kernel-parameters.txt +6 -0
index 8228c9acc530..ed505e362b1d 100644
@@ -424,6 +424,12 @@ Kernel parameters
 			do not want to use tracing_snapshot_alloc() as it needs
 			to be done where GFP_KERNEL allocations are allowed.
 
+	allow_driver_override
+			Allow matching drivers without
+			.support_driver_override = true to devices using
+			driver_override. This is a dangerous operation in
+			general allowing root to oops the kernel.
+
 	allow_mismatched_32bit_el0 [ARM64,EARLY]
 			Allow execve() of 32-bit applications and setting of the
 			PER_LINUX32 personality on systems where only a strict
Documentation/admin-guide/tainted-kernels.rst +29 -25

67 lines not on this page · raw message

drivers/base/bus.c +51 -0

57 lines not on this page · raw message

include/linux/device.h +2 -17
index 6b082c638d45..05413c1a9454 100644
@@ -907,23 +907,8 @@ static inline bool device_has_driver_override(struct device *dev)
 	return !!dev->driver_override.name;
 }
 
-/**
- * device_match_driver_override() - Match a driver against the device's driver_override.
- * @dev: device to check
- * @drv: driver to match against
- *
- * Returns > 0 if a driver override is set and matches the given driver, 0 if a
- * driver override is set but does not match, or < 0 if a driver override is not
- * set at all.
- */
-static inline int device_match_driver_override(struct device *dev,
-					       const struct device_driver *drv)
-{
-	guard(spinlock)(&dev->driver_override.lock);
-	if (dev->driver_override.name)
-		return !strcmp(dev->driver_override.name, drv->name);
-	return -1;
-}
+int device_match_driver_override(struct device *dev,
+				 const struct device_driver *drv);
 
 /**
  * device_iommu_mapped - Returns true when the device DMA is translated
include/linux/device/driver.h +2 -0
index 29fbc01ef06f..0153cfcbe1b9 100644
@@ -56,6 +56,7 @@ enum probe_type {
  * @bus:	The bus which the device of this driver belongs to.
  * @owner:	The module owner.
  * @mod_name:	Used for built-in modules.
+ * @support_driver_override: driver_override only works if this is true.
  * @suppress_bind_attrs: Disables bind/unbind via sysfs.
  * @probe_type:	Type of the probe (synchronous or asynchronous) to use.
  * @of_match_table: The open firmware table.
@@ -104,6 +105,7 @@ struct device_driver {
 	struct module		*owner;
 	const char		*mod_name;	/* used for built-in modules */
 
+	bool support_driver_override;
 	bool suppress_bind_attrs;	/* disables bind/unbind via sysfs */
 	enum probe_type probe_type;
 
include/linux/panic.h +2 -1
index 17e61b61c45f..009232511cc7 100644
@@ -89,7 +89,8 @@ static inline void set_arch_panic_timeout(int timeout, int arch_default_timeout)
 #define TAINT_RANDSTRUCT		17
 #define TAINT_TEST			18
 #define TAINT_FWCTL			19
-#define TAINT_FLAGS_COUNT		20
+#define TAINT_DRIVER_OVERRIDE		20
+#define TAINT_FLAGS_COUNT		21
 #define TAINT_FLAGS_MAX			((1UL << TAINT_FLAGS_COUNT) - 1)
 
 struct taint_flag {
include/trace/events/module.h +2 -1
index e5a006be9dc6..d54ce3c7572b 100644
@@ -26,7 +26,8 @@ struct module;
 	{ (1UL << TAINT_OOT_MODULE),		"O" },		\
 	{ (1UL << TAINT_FORCED_MODULE),		"F" },		\
 	{ (1UL << TAINT_CRAP),			"C" },		\
-	{ (1UL << TAINT_UNSIGNED_MODULE),	"E" })
+	{ (1UL << TAINT_UNSIGNED_MODULE),	"E" },		\
+	{ (1UL << TAINT_DRIVER_OVERRIDE),	"Z" })
 
 TRACE_EVENT(module_load,
 
kernel/panic.c +2 -1
index 91e48d33b7db..81ea9ef9afb0 100644
@@ -825,6 +825,7 @@ const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
 	TAINT_FLAG(RANDSTRUCT,			'T', ' '),
 	TAINT_FLAG(TEST,			'N', ' '),
 	TAINT_FLAG(FWCTL,			'J', ' '),
+	TAINT_FLAG(DRIVER_OVERRIDE,		'Z', ' '),
 };
 
 #undef TAINT_FLAG
@@ -861,7 +862,7 @@ static void print_tainted_seq(struct seq_buf *s, bool verbose)
  * exact size is allocated dynamically; the initial buffer remains
  * as a fallback if allocation fails.
  *
- * The verbose taint string currently requires up to 327 characters.
+ * The verbose taint string currently requires up to 348 characters.
  */
 #define INIT_TAINT_BUF_MAX 350
 
tools/debugging/kernel-chktaint +9 -0
index e1571c04afb5..56215e0f5112 100755
@@ -211,6 +211,15 @@ else
 	addout "J"
 	echo " * fwctl's mutating debug interface was used (#19)"
 fi
+
+T=`expr $T / 2`
+if [ `expr $T % 2` -eq 0 ]; then
+	addout " "
+else
+	addout "Z"
+	echo " * A device was matched using driver_override to a driver not supporting that (#20)"
+fi
+
 echo "Raw taint value as int/string: $taint/'$out'"
 
 # report on any tainted loadable modules
Signature · 1 line
2.56.0.rc1
#2 lore raw
Quoted 5 lines · Uwe Kleine-König
Documentation/admin-guide/kernel-parameters.txt +0 -0
index 8228c9acc530..ed505e362b1d 100644

Documentation/admin-guide/kernel-parameters.txt Kernel parameters · line 432

Quoted 9 lines
Documentation/admin-guide/kernel-parameters.txt +6 -0
@@ -424,6 +424,12 @@ Kernel parameters
 			do not want to use tracing_snapshot_alloc() as it needs
 			to be done where GFP_KERNEL allocations are allowed.
 
+	allow_driver_override
+			Allow matching drivers without
+			.support_driver_override = true to devices using
+			driver_override. This is a dangerous operation in
+			general allowing root to oops the kernel.
+

I feel like this is more appropriate as a lockdown level than a kernel command line?

Jason

Reviewed-by Tested-by on v2 1/10
#3 lore raw
In reply to Uwe Kleine-König

From: Uwe Kleine-König <[email protected]> Sent: Friday, November 20, 2026 3:11 AM

Quoted 7 lines

Driver overriding is useful only in a very limited set of situations and then only with very few drivers that are designed to support that.

Disallow matching via driver_override unless the driver explicitly allows it or the safe guard is disabled using the "allow_driver_override=1" kernel parameter.

The "=1" syntax is no longer accurate.

Quoted 5 lines

There are many drivers that don't expect to be force-bound and often doing it anyhow results in a NULL pointer exception. To make debugging such cases easier, add a new taint if such a forced match happens.

Quoted 7 lines

Implementation detail: device_match_driver_override() isn't a static inline any more. It grew a certain complexity (e.g. a pr_info()) and so was made a regular exported function.

Modulo the nit in the commit message,

Quoted 288 lines
---
 .../admin-guide/kernel-parameters.txt         |  6 +++
 Documentation/admin-guide/tainted-kernels.rst | 54 ++++++++++---------
 drivers/base/bus.c                            | 51 ++++++++++++++++++
 include/linux/device.h                        | 19 +------
 include/linux/device/driver.h                 |  2 +
 include/linux/panic.h                         |  3 +-
 include/trace/events/module.h                 |  3 +-
 kernel/panic.c                                |  3 +-
 tools/debugging/kernel-chktaint               |  9 ++++
 9 files changed, 105 insertions(+), 45 deletions(-)
Documentation/admin-guide/kernel-parameters.txt +6 -0
index 8228c9acc530..ed505e362b1d 100644
@@ -424,6 +424,12 @@ Kernel parameters
 			do not want to use tracing_snapshot_alloc() as it needs
 			to be done where GFP_KERNEL allocations are allowed.
 
+	allow_driver_override
+			Allow matching drivers without
+			.support_driver_override = true to devices using
+			driver_override. This is a dangerous operation in
+			general allowing root to oops the kernel.
+
 	allow_mismatched_32bit_el0 [ARM64,EARLY]
 			Allow execve() of 32-bit applications and setting of the
 			PER_LINUX32 personality on systems where only a strict
Documentation/admin-guide/tainted-kernels.rst +29 -25
index fcb8b9734548..f4ddb63dc704 100644
@@ -74,35 +74,36 @@ a particular type of taint. It's best to leave that to the aforementioned
 script, but if you need something quick you can use this shell command to check
 which bits are set::
 
-	$ for i in $(seq 0 19); do echo $i $(($(cat /proc/sys/kernel/tainted)>>$i&1));done
+	$ for i in $(seq 0 20); do echo $i $(($(cat /proc/sys/kernel/tainted)>>$i&1));done
 
 Table for decoding tainted state
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-===  ===  ======  ========================================================
-Bit  Log  Number  Reason that got the kernel tainted
-===  ===  ======  ========================================================
-  0  G/P       1  proprietary module was loaded
-  1  _/F       2  module was force loaded
-  2  _/S       4  kernel running on an out of specification system
-  3  _/R       8  module was force unloaded
-  4  _/M      16  processor reported a Machine Check Exception (MCE)
-  5  _/B      32  bad page referenced or some unexpected page flags
-  6  _/U      64  taint requested by userspace application
-  7  _/D     128  kernel died recently, i.e. there was an OOPS or BUG
-  8  _/A     256  ACPI table overridden by user
-  9  _/W     512  kernel issued warning
- 10  _/C    1024  staging driver was loaded
- 11  _/I    2048  workaround for bug in platform firmware applied
- 12  _/O    4096  externally-built ("out-of-tree") module was loaded
- 13  _/E    8192  unsigned module was loaded
- 14  _/L   16384  soft lockup occurred
- 15  _/K   32768  kernel has been live patched
- 16  _/X   65536  auxiliary taint, defined for and used by distros
- 17  _/T  131072  kernel was built with the struct randomization plugin
- 18  _/N  262144  an in-kernel test has been run
- 19  _/J  524288  userspace used a mutating debug operation in fwctl
-===  ===  ======  ========================================================
+===  ===  =======  ========================================================
+Bit  Log   Number  Reason that got the kernel tainted
+===  ===  =======  ========================================================
+  0  G/P        1  proprietary module was loaded
+  1  _/F        2  module was force loaded
+  2  _/S        4  kernel running on an out of specification system
+  3  _/R        8  module was force unloaded
+  4  _/M       16  processor reported a Machine Check Exception (MCE)
+  5  _/B       32  bad page referenced or some unexpected page flags
+  6  _/U       64  taint requested by userspace application
+  7  _/D      128  kernel died recently, i.e. there was an OOPS or BUG
+  8  _/A      256  ACPI table overridden by user
+  9  _/W      512  kernel issued warning
+ 10  _/C     1024  staging driver was loaded
+ 11  _/I     2048  workaround for bug in platform firmware applied
+ 12  _/O     4096  externally-built ("out-of-tree") module was loaded
+ 13  _/E     8192  unsigned module was loaded
+ 14  _/L    16384  soft lockup occurred
+ 15  _/K    32768  kernel has been live patched
+ 16  _/X    65536  auxiliary taint, defined for and used by distros
+ 17  _/T   131072  kernel was built with the struct randomization plugin
+ 18  _/N   262144  an in-kernel test has been run
+ 19  _/J   524288  userspace used a mutating debug operation in fwctl
+ 20  _/Z  1048576  A device was matched using driver_override to a driver not supporting that
+===  ===  =======  ========================================================
 
 Note: The character ``_`` is representing a blank in this table to make reading
 easier.
@@ -189,3 +190,6 @@ More detailed explanation for tainting
  19) ``J`` if userspace opened /dev/fwctl/* and performed a FWTCL_RPC_DEBUG_WRITE
      to use the devices debugging features. Device debugging features could
      cause the device to malfunction in undefined ways.
+
+ 20) ``Z`` if a device was matched using the `driver_override` sysfs property
+     to a driver that didn't explicitly allow that.
drivers/base/bus.c +51 -0
index 396097fdb316..1b4dd98a1286 100644
@@ -603,6 +603,57 @@ int bus_add_device(struct device *dev)
 	return error;
 }
 
+static bool __read_mostly allow_driver_override;
+
+static int __init allow_driver_override_setup(char *str)
+{
+	if (!str || *str != '\0')
+		return 0;
+
+	allow_driver_override = true;
+	return 1;
+}
+__setup("allow_driver_override", allow_driver_override_setup);
+
+/**
+ * device_match_driver_override() - Match a driver against the device's driver_override.
+ * @dev: device to check
+ * @drv: driver to match against
+ *
+ * Returns > 0 if a driver override is set and matches the given driver, 0 if a
+ * driver override is set but does not match or is ignored, or < 0 if a driver
+ * override is not set at all.
+ */
+int device_match_driver_override(struct device *dev,
+				 const struct device_driver *drv)
+{
+	guard(spinlock)(&dev->driver_override.lock);
+	if (dev->driver_override.name) {
+		int ret = !strcmp(dev->driver_override.name, drv->name);
+
+		if (ret && !drv->support_driver_override) {
+			if (allow_driver_override) {
+				add_taint_module(drv->owner,
+						 TAINT_DRIVER_OVERRIDE, LOCKDEP_STILL_OK);
+			} else {
+				/*
+				 * If you hit this warning and you need to
+				 * quickly workaround this check, boot with
+				 * allow_driver_override=1 on the kernel
+				 * commandline. The right fix is to set the
+				 * driver's .support_driver_override property.
+				 */
+				dev_warn(dev, "Denied matching to %s via driver override.\n", drv->name);
+				return 0;
+			}
+		}
+
+		return ret;
+	}
+	return -1;
+}
+EXPORT_SYMBOL_GPL(device_match_driver_override);
+
 static void __bus_probe_device(struct device *dev)
 {
 	struct subsys_private *sp = bus_to_subsys(dev->bus);
include/linux/device.h +2 -17
index 6b082c638d45..05413c1a9454 100644
@@ -907,23 +907,8 @@ static inline bool device_has_driver_override(struct device *dev)
 	return !!dev->driver_override.name;
 }
 
-/**
- * device_match_driver_override() - Match a driver against the device's driver_override.
- * @dev: device to check
- * @drv: driver to match against
- *
- * Returns > 0 if a driver override is set and matches the given driver, 0 if a
- * driver override is set but does not match, or < 0 if a driver override is not
- * set at all.
- */
-static inline int device_match_driver_override(struct device *dev,
-					       const struct device_driver *drv)
-{
-	guard(spinlock)(&dev->driver_override.lock);
-	if (dev->driver_override.name)
-		return !strcmp(dev->driver_override.name, drv->name);
-	return -1;
-}
+int device_match_driver_override(struct device *dev,
+				 const struct device_driver *drv);
 
 /**
  * device_iommu_mapped - Returns true when the device DMA is translated
include/linux/device/driver.h +2 -0
index 29fbc01ef06f..0153cfcbe1b9 100644
@@ -56,6 +56,7 @@ enum probe_type {
  * @bus:	The bus which the device of this driver belongs to.
  * @owner:	The module owner.
  * @mod_name:	Used for built-in modules.
+ * @support_driver_override: driver_override only works if this is true.
  * @suppress_bind_attrs: Disables bind/unbind via sysfs.
  * @probe_type:	Type of the probe (synchronous or asynchronous) to use.
  * @of_match_table: The open firmware table.
@@ -104,6 +105,7 @@ struct device_driver {
 	struct module		*owner;
 	const char		*mod_name;	/* used for built-in modules */
 
+	bool support_driver_override;
 	bool suppress_bind_attrs;	/* disables bind/unbind via sysfs */
 	enum probe_type probe_type;
 
include/linux/panic.h +2 -1
index 17e61b61c45f..009232511cc7 100644
@@ -89,7 +89,8 @@ static inline void set_arch_panic_timeout(int timeout, int
 arch_default_timeout)
 #define TAINT_RANDSTRUCT		17
 #define TAINT_TEST			18
 #define TAINT_FWCTL			19
-#define TAINT_FLAGS_COUNT		20
+#define TAINT_DRIVER_OVERRIDE		20
+#define TAINT_FLAGS_COUNT		21
 #define TAINT_FLAGS_MAX			((1UL << TAINT_FLAGS_COUNT) - 1)
 

struct taint_flag {

include/trace/events/module.h +2 -1
index e5a006be9dc6..d54ce3c7572b 100644
@@ -26,7 +26,8 @@ struct module;
 	{ (1UL << TAINT_OOT_MODULE),		"O" },		\
 	{ (1UL << TAINT_FORCED_MODULE),		"F" },		\
 	{ (1UL << TAINT_CRAP),			"C" },		\
-	{ (1UL << TAINT_UNSIGNED_MODULE),	"E" })
+	{ (1UL << TAINT_UNSIGNED_MODULE),	"E" },		\
+	{ (1UL << TAINT_DRIVER_OVERRIDE),	"Z" })
 
 TRACE_EVENT(module_load,
 
kernel/panic.c +2 -1
index 91e48d33b7db..81ea9ef9afb0 100644
@@ -825,6 +825,7 @@ const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
 	TAINT_FLAG(RANDSTRUCT,			'T', ' '),
 	TAINT_FLAG(TEST,			'N', ' '),
 	TAINT_FLAG(FWCTL,			'J', ' '),
+	TAINT_FLAG(DRIVER_OVERRIDE,		'Z', ' '),
 };
 
 #undef TAINT_FLAG
@@ -861,7 +862,7 @@ static void print_tainted_seq(struct seq_buf *s, bool verbose)
  * exact size is allocated dynamically; the initial buffer remains
  * as a fallback if allocation fails.
  *
- * The verbose taint string currently requires up to 327 characters.
+ * The verbose taint string currently requires up to 348 characters.
  */
 #define INIT_TAINT_BUF_MAX 350
 
tools/debugging/kernel-chktaint +9 -1
index e1571c04afb5..56215e0f5112 100755
@@ -211,6 +211,15 @@ else
 	addout "J"
 	echo " * fwctl's mutating debug interface was used (#19)"
 fi
+
+T=`expr $T / 2`
+if [ `expr $T % 2` -eq 0 ]; then
+	addout " "
+else
+	addout "Z"
+	echo " * A device was matched using driver_override to a driver not supporting that (#20)"
+fi
+
 echo "Raw taint value as int/string: $taint/'$out'"
 
 # report on any tainted loadable modules
--

2.56.0.rc1

#4 lore raw

PATCH v2 2/10 PCI: Allow untainted driver_override for PCI stub drivers

In reply to Uwe Kleine-König

From: Alex Williamson <[email protected]>

As the name suggests, the pci-stub driver is only a stub, it performs no hardware access and is typically used only to prevent other drivers from binding to the device.

Similarly, pci-pf-stub only interacts with devices through the SR-IOV capability, and may be used to create VFs on any PF that otherwise does not require a PF driver.

Both of these are often used with driver_override, as this is our standard, non-racy mechanism for associating a driver to a device. With the introduction of support_driver_override in driver core, this use case needs to be explicitly flagged as safe or else the bind will be denied or taint the kernel. Declare these drivers as safe for use with driver_override.

---
 drivers/pci/pci-pf-stub.c | 1 +
 drivers/pci/pci-stub.c    | 1 +
 2 files changed, 2 insertions(+)
drivers/pci/pci-pf-stub.c +1 -0
index da4db4928907..328fb4a57092 100644
@@ -36,6 +36,7 @@ static struct pci_driver pf_stub_driver = {
 	.id_table		= pci_pf_stub_whitelist,
 	.probe			= pci_pf_stub_probe,
 	.sriov_configure	= pci_sriov_configure_simple,
+	.driver.support_driver_override = true,
 };
 module_pci_driver(pf_stub_driver);
 
drivers/pci/pci-stub.c +1 -0
index 9bc478df4e8f..acbd43c17adc 100644
@@ -37,6 +37,7 @@ static struct pci_driver stub_driver = {
 	.id_table	= NULL,	/* only dynamic id's */
 	.probe		= pci_stub_probe,
 	.driver_managed_dma = true,
+	.driver.support_driver_override = true,
 };
 
 static int __init pci_stub_init(void)
Signature · 1 line
2.56.0.rc1
#5 lore raw

PATCH v2 3/10 vfio: Allow untainted use of driver_override

In reply to Uwe Kleine-König

From: Alex Williamson <[email protected]>

Driver overrides are the standard mechanism for binding vfio drivers to devices. With the introduction of the support_driver_override flag in the core driver struct, it's now required for drivers to declare themselves as driver_override compatible, otherwise driver_override bindings will be denied or generate a kernel taint.

Set the support_driver_override flag on the driver struct across all vfio drivers.

---
 drivers/vfio/cdx/main.c                        | 1 +
 drivers/vfio/fsl-mc/vfio_fsl_mc.c              | 1 +
 drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 3 ++-
 drivers/vfio/pci/ism/main.c                    | 3 ++-
 drivers/vfio/pci/mlx5/main.c                   | 3 ++-
 drivers/vfio/pci/nvgrace-gpu/main.c            | 3 ++-
 drivers/vfio/pci/pds/pci_drv.c                 | 3 ++-
 drivers/vfio/pci/qat/main.c                    | 3 ++-
 drivers/vfio/pci/vfio_pci.c                    | 3 ++-
 drivers/vfio/pci/virtio/main.c                 | 3 ++-
 drivers/vfio/pci/xe/main.c                     | 3 ++-
 drivers/vfio/platform/vfio_amba.c              | 1 +
 drivers/vfio/platform/vfio_platform.c          | 1 +
 13 files changed, 22 insertions(+), 9 deletions(-)
drivers/vfio/cdx/main.c +1 -0
index 28dc8c825600..fb7679bf92fd 100644
@@ -358,6 +358,7 @@ static struct cdx_driver vfio_cdx_driver = {
 	.match_id_table	= vfio_cdx_table,
 	.driver	= {
 		.name	= "vfio-cdx",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
drivers/vfio/fsl-mc/vfio_fsl_mc.c +1 -0
index 15b05b8c0c90..1603f81e0257 100644
@@ -597,6 +597,7 @@ static struct fsl_mc_driver vfio_fsl_mc_driver = {
 	.remove		= vfio_fsl_mc_remove,
 	.driver	= {
 		.name	= "vfio-fsl-mc",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +2 -1
index c82c59246a3f..5c56aff8db01 100644
@@ -1737,7 +1737,8 @@ static struct pci_driver hisi_acc_vfio_pci_driver = {
 	.id_table = hisi_acc_vfio_pci_table,
 	.probe = hisi_acc_vfio_pci_probe,
 	.remove = hisi_acc_vfio_pci_remove,
-	.driver = { .pm = pm_ptr(&hisi_acc_vf_pm_ops) },
+	.driver = { .pm = pm_ptr(&hisi_acc_vf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &hisi_acc_vf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/ism/main.c +2 -1
index cce3bdf1f070..263b377e87c4 100644
@@ -399,7 +399,8 @@ static struct pci_driver ism_vfio_pci_driver = {
 	.id_table = ism_device_table,
 	.probe = ism_vfio_pci_probe,
 	.remove = ism_vfio_pci_remove,
-	.driver	= { .pm = pm_ptr(&ism_vfio_pm_ops) },
+	.driver	= { .pm = pm_ptr(&ism_vfio_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &ism_vfio_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/mlx5/main.c +2 -1
index 18a6d224b1fd..63c89e461be8 100644
@@ -1451,7 +1451,8 @@ static struct pci_driver mlx5vf_pci_driver = {
 	.id_table = mlx5vf_pci_table,
 	.probe = mlx5vf_pci_probe,
 	.remove = mlx5vf_pci_remove,
-	.driver	= { .pm = pm_ptr(&mlx5vf_pm_ops) },
+	.driver	= { .pm = pm_ptr(&mlx5vf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &mlx5vf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/nvgrace-gpu/main.c +2 -1
index 5ddec340b73b..946da32c61ff 100644
@@ -1438,7 +1438,8 @@ static struct pci_driver nvgrace_gpu_vfio_pci_driver = {
 	.id_table = nvgrace_gpu_vfio_pci_table,
 	.probe = nvgrace_gpu_probe,
 	.remove = nvgrace_gpu_remove,
-	.driver = { .pm = pm_ptr(&nvgrace_gpu_pm_ops) },
+	.driver = { .pm = pm_ptr(&nvgrace_gpu_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &nvgrace_gpu_vfio_pci_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/pds/pci_drv.c +2 -1
index cf7e33d4d17b..1fcef987f1b0 100644
@@ -184,7 +184,8 @@ static struct pci_driver pds_vfio_pci_driver = {
 	.id_table = pds_vfio_pci_table,
 	.probe = pds_vfio_pci_probe,
 	.remove = pds_vfio_pci_remove,
-	.driver = { .pm = pm_ptr(&pds_vfio_pm_ops) },
+	.driver = { .pm = pm_ptr(&pds_vfio_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &pds_vfio_pci_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/qat/main.c +2 -1
index 69448fa57df4..8a616f9b6494 100644
@@ -695,7 +695,8 @@ static struct pci_driver qat_vf_vfio_pci_driver = {
 	.id_table = qat_vf_vfio_pci_table,
 	.probe = qat_vf_vfio_pci_probe,
 	.remove = qat_vf_vfio_pci_remove,
-	.driver	= { .pm = pm_ptr(&qat_vf_pm_ops) },
+	.driver	= { .pm = pm_ptr(&qat_vf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &qat_vf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/vfio_pci.c +2 -1
index e7c012febe91..fde2c3fc88eb 100644
@@ -260,7 +260,8 @@ static struct pci_driver vfio_pci_driver = {
 	.remove			= vfio_pci_remove,
 	.sriov_configure	= vfio_pci_sriov_configure,
 	.err_handler		= &vfio_pci_err_handlers,
-	.driver		= { .pm = pm_ptr(&vfio_pci_pm_ops) },
+	.driver		= { .pm = pm_ptr(&vfio_pci_pm_ops),
+			    .support_driver_override = true },
 	.driver_managed_dma	= true,
 };
 
drivers/vfio/pci/virtio/main.c +2 -1
index 123ff5193812..1e02a8fd4a61 100644
@@ -222,7 +222,8 @@ static struct pci_driver virtiovf_pci_driver = {
 	.id_table = virtiovf_pci_table,
 	.probe = virtiovf_pci_probe,
 	.remove = virtiovf_pci_remove,
-	.driver = { .pm = pm_ptr(&virtiovf_pm_ops) },
+	.driver = { .pm = pm_ptr(&virtiovf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &virtiovf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/xe/main.c +2 -1
index 7dab3e024aa8..e0ab61f604de 100644
@@ -592,7 +592,8 @@ static struct pci_driver xe_vfio_pci_driver = {
 	.id_table = xe_vfio_pci_table,
 	.probe = xe_vfio_pci_probe,
 	.remove = xe_vfio_pci_remove,
-	.driver = { .pm = pm_ptr(&xe_vfio_pm_ops) },
+	.driver = { .pm = pm_ptr(&xe_vfio_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &xe_vfio_pci_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/platform/vfio_amba.c +1 -0
index fa754f203b2d..2c87caf1b7f7 100644
@@ -137,6 +137,7 @@ static struct amba_driver vfio_amba_driver = {
 	.id_table = vfio_amba_ids,
 	.drv = {
 		.name = "vfio-amba",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
drivers/vfio/platform/vfio_platform.c +1 -0
index a4d3ace3e02d..93d06032cdde 100644
@@ -116,6 +116,7 @@ static struct platform_driver vfio_platform_driver = {
 	.remove		= vfio_platform_remove,
 	.driver	= {
 		.name	= "vfio-platform",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
Signature · 1 line
2.56.0.rc1
Reviewed-by on v2 3/10
#6 lore raw

Re: [PATCH v2 03/10] vfio: Allow untainted use of driver_override

Quoted 8 lines · Uwe Kleine-König

From: Alex Williamson <[email protected]>

Driver overrides are the standard mechanism for binding vfio drivers to devices. With the introduction of the support_driver_override flag in the core driver struct, it's now required for drivers to declare themselves as driver_override compatible, otherwise driver_override bindings will be denied or generate a kernel taint.

Quoted 8 lines

Set the support_driver_override flag on the driver struct across all vfio drivers.

LGTM.

Quoted 194 lines
---
 drivers/vfio/cdx/main.c                        | 1 +
 drivers/vfio/fsl-mc/vfio_fsl_mc.c              | 1 +
 drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 3 ++-
 drivers/vfio/pci/ism/main.c                    | 3 ++-
 drivers/vfio/pci/mlx5/main.c                   | 3 ++-
 drivers/vfio/pci/nvgrace-gpu/main.c            | 3 ++-
 drivers/vfio/pci/pds/pci_drv.c                 | 3 ++-
 drivers/vfio/pci/qat/main.c                    | 3 ++-
 drivers/vfio/pci/vfio_pci.c                    | 3 ++-
 drivers/vfio/pci/virtio/main.c                 | 3 ++-
 drivers/vfio/pci/xe/main.c                     | 3 ++-
 drivers/vfio/platform/vfio_amba.c              | 1 +
 drivers/vfio/platform/vfio_platform.c          | 1 +
 13 files changed, 22 insertions(+), 9 deletions(-)
drivers/vfio/cdx/main.c +1 -0
index 28dc8c825600..fb7679bf92fd 100644
@@ -358,6 +358,7 @@ static struct cdx_driver vfio_cdx_driver = {
 	.match_id_table	= vfio_cdx_table,
 	.driver	= {
 		.name	= "vfio-cdx",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
drivers/vfio/fsl-mc/vfio_fsl_mc.c +1 -0
index 15b05b8c0c90..1603f81e0257 100644
@@ -597,6 +597,7 @@ static struct fsl_mc_driver vfio_fsl_mc_driver = {
 	.remove		= vfio_fsl_mc_remove,
 	.driver	= {
 		.name	= "vfio-fsl-mc",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +2 -1
index c82c59246a3f..5c56aff8db01 100644
@@ -1737,7 +1737,8 @@ static struct pci_driver hisi_acc_vfio_pci_driver = {
 	.id_table = hisi_acc_vfio_pci_table,
 	.probe = hisi_acc_vfio_pci_probe,
 	.remove = hisi_acc_vfio_pci_remove,
-	.driver = { .pm = pm_ptr(&hisi_acc_vf_pm_ops) },
+	.driver = { .pm = pm_ptr(&hisi_acc_vf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &hisi_acc_vf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/ism/main.c +2 -1
index cce3bdf1f070..263b377e87c4 100644
@@ -399,7 +399,8 @@ static struct pci_driver ism_vfio_pci_driver = {
 	.id_table = ism_device_table,
 	.probe = ism_vfio_pci_probe,
 	.remove = ism_vfio_pci_remove,
-	.driver	= { .pm = pm_ptr(&ism_vfio_pm_ops) },
+	.driver	= { .pm = pm_ptr(&ism_vfio_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &ism_vfio_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/mlx5/main.c +2 -1
index 18a6d224b1fd..63c89e461be8 100644
@@ -1451,7 +1451,8 @@ static struct pci_driver mlx5vf_pci_driver = {
 	.id_table = mlx5vf_pci_table,
 	.probe = mlx5vf_pci_probe,
 	.remove = mlx5vf_pci_remove,
-	.driver	= { .pm = pm_ptr(&mlx5vf_pm_ops) },
+	.driver	= { .pm = pm_ptr(&mlx5vf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &mlx5vf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/nvgrace-gpu/main.c +2 -1
index 5ddec340b73b..946da32c61ff 100644
@@ -1438,7 +1438,8 @@ static struct pci_driver nvgrace_gpu_vfio_pci_driver = {
 	.id_table = nvgrace_gpu_vfio_pci_table,
 	.probe = nvgrace_gpu_probe,
 	.remove = nvgrace_gpu_remove,
-	.driver = { .pm = pm_ptr(&nvgrace_gpu_pm_ops) },
+	.driver = { .pm = pm_ptr(&nvgrace_gpu_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &nvgrace_gpu_vfio_pci_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/pds/pci_drv.c +2 -1
index cf7e33d4d17b..1fcef987f1b0 100644
@@ -184,7 +184,8 @@ static struct pci_driver pds_vfio_pci_driver = {
 	.id_table = pds_vfio_pci_table,
 	.probe = pds_vfio_pci_probe,
 	.remove = pds_vfio_pci_remove,
-	.driver = { .pm = pm_ptr(&pds_vfio_pm_ops) },
+	.driver = { .pm = pm_ptr(&pds_vfio_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &pds_vfio_pci_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/qat/main.c +2 -1
index 69448fa57df4..8a616f9b6494 100644
@@ -695,7 +695,8 @@ static struct pci_driver qat_vf_vfio_pci_driver = {
 	.id_table = qat_vf_vfio_pci_table,
 	.probe = qat_vf_vfio_pci_probe,
 	.remove = qat_vf_vfio_pci_remove,
-	.driver	= { .pm = pm_ptr(&qat_vf_pm_ops) },
+	.driver	= { .pm = pm_ptr(&qat_vf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &qat_vf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/vfio_pci.c +2 -1
index e7c012febe91..fde2c3fc88eb 100644
@@ -260,7 +260,8 @@ static struct pci_driver vfio_pci_driver = {
 	.remove			= vfio_pci_remove,
 	.sriov_configure	= vfio_pci_sriov_configure,
 	.err_handler		= &vfio_pci_err_handlers,
-	.driver		= { .pm = pm_ptr(&vfio_pci_pm_ops) },
+	.driver		= { .pm = pm_ptr(&vfio_pci_pm_ops),
+			    .support_driver_override = true },
 	.driver_managed_dma	= true,
 };
 
drivers/vfio/pci/virtio/main.c +2 -1
index 123ff5193812..1e02a8fd4a61 100644
@@ -222,7 +222,8 @@ static struct pci_driver virtiovf_pci_driver = {
 	.id_table = virtiovf_pci_table,
 	.probe = virtiovf_pci_probe,
 	.remove = virtiovf_pci_remove,
-	.driver = { .pm = pm_ptr(&virtiovf_pm_ops) },
+	.driver = { .pm = pm_ptr(&virtiovf_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &virtiovf_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/pci/xe/main.c +2 -1
index 7dab3e024aa8..e0ab61f604de 100644
@@ -592,7 +592,8 @@ static struct pci_driver xe_vfio_pci_driver = {
 	.id_table = xe_vfio_pci_table,
 	.probe = xe_vfio_pci_probe,
 	.remove = xe_vfio_pci_remove,
-	.driver = { .pm = pm_ptr(&xe_vfio_pm_ops) },
+	.driver = { .pm = pm_ptr(&xe_vfio_pm_ops),
+		    .support_driver_override = true },
 	.err_handler = &xe_vfio_pci_err_handlers,
 	.driver_managed_dma = true,
 };
drivers/vfio/platform/vfio_amba.c +1 -0
index fa754f203b2d..2c87caf1b7f7 100644
@@ -137,6 +137,7 @@ static struct amba_driver vfio_amba_driver = {
 	.id_table = vfio_amba_ids,
 	.drv = {
 		.name = "vfio-amba",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
drivers/vfio/platform/vfio_platform.c +1 -1
index a4d3ace3e02d..93d06032cdde 100644
@@ -116,6 +116,7 @@ static struct platform_driver vfio_platform_driver = {
 	.remove		= vfio_platform_remove,
 	.driver	= {
 		.name	= "vfio-platform",
+		.support_driver_override = true,
 	},
 	.driver_managed_dma = true,
 };
--

2.56.0.rc1

#7 lore raw

PATCH v2 4/10 s390/vfio: Allow untainted use of driver_override

In reply to Uwe Kleine-König

From: Alex Williamson <[email protected]>

The vfio_ap and vfio_ccw drivers bind devices through driver_override. With the introduction of the support_driver_override flag in the core driver struct, it's now required for drivers to declare themselves as driver_override compatible, otherwise driver_override bindings will be denied or generate a kernel taint.

Set the support_driver_override flag on the vfio_ap and vfio_ccw driver structs.

---
 drivers/s390/cio/vfio_ccw_drv.c   | 1 +
 drivers/s390/crypto/vfio_ap_drv.c | 1 +
 2 files changed, 2 insertions(+)
drivers/s390/cio/vfio_ccw_drv.c +1 -0
index 7546bd137e4e..77765cde87f7 100644
@@ -375,6 +375,7 @@ static struct css_driver vfio_ccw_sch_driver = {
 	.drv = {
 		.name = "vfio_ccw",
 		.owner = THIS_MODULE,
+		.support_driver_override = true,
 	},
 	.subchannel_type = vfio_ccw_sch_ids,
 	.irq = vfio_ccw_sch_irq,
drivers/s390/crypto/vfio_ap_drv.c +1 -0
index 8e69ed286bb9..2d47dc643269 100644
@@ -62,6 +62,7 @@ static struct ap_driver vfio_ap_drv = {
 	.on_config_changed = vfio_ap_on_cfg_changed,
 	.on_scan_complete = vfio_ap_on_scan_complete,
 	.ids = ap_queue_ids,
+	.driver.support_driver_override = true,
 };
 
 static void vfio_ap_matrix_dev_release(struct device *dev)
Signature · 1 line
2.56.0.rc1
#8 lore raw

Re: [PATCH v2 04/10] s390/vfio: Allow untainted use of driver_override

Quoted 36 lines · Uwe Kleine-König

From: Alex Williamson <[email protected]>

The vfio_ap and vfio_ccw drivers bind devices through driver_override. With the introduction of the support_driver_override flag in the core driver struct, it's now required for drivers to declare themselves as driver_override compatible, otherwise driver_override bindings will be denied or generate a kernel taint.

Set the support_driver_override flag on the vfio_ap and vfio_ccw driver structs.

---
 drivers/s390/cio/vfio_ccw_drv.c   | 1 +
 drivers/s390/crypto/vfio_ap_drv.c | 1 +
 2 files changed, 2 insertions(+)
drivers/s390/cio/vfio_ccw_drv.c +1 -0
index 7546bd137e4e..77765cde87f7 100644
@@ -375,6 +375,7 @@ static struct css_driver vfio_ccw_sch_driver = {
 	.drv = {
 		.name = "vfio_ccw",
 		.owner = THIS_MODULE,
+		.support_driver_override = true,
 	},
 	.subchannel_type = vfio_ccw_sch_ids,
 	.irq = vfio_ccw_sch_irq,
drivers/s390/crypto/vfio_ap_drv.c +0 -0
index 8e69ed286bb9..2d47dc643269 100644

drivers/s390/crypto/vfio_ap_drv.c static struct ap_driver vfio_ap_drv = { · line 68

Quoted 7 lines
drivers/s390/crypto/vfio_ap_drv.c +1 -0
@@ -62,6 +62,7 @@ static struct ap_driver vfio_ap_drv = {
 	.on_config_changed = vfio_ap_on_cfg_changed,
 	.on_scan_complete = vfio_ap_on_scan_complete,
 	.ids = ap_queue_ids,
+	.driver.support_driver_override = true,
 };
 
 static void vfio_ap_matrix_dev_release(struct device *dev)

Hm, honestly I don' get this...
The driver_override is a bus thing. So the AP bus has support for driver_override. The sysfs functions are in drivers/s390/crypto/ap_queue.c. There exist currently two drivers which shall be bound to an AP queue: vfio_ap and the default driver which is the zcrypt driver "cex4queue". It sounds like the driver needs the tagging about driver_override support and then similar to the vfio_ap_drv.c the static struct ap_driver initialization may also need a .driver.support_driver_override = true line.

regards
Harald Freudenberger

#9 lore raw

Re: [PATCH v2 04/10] s390/vfio: Allow untainted use of driver_override

Quoted 44 lines · Harald Freudenberger
Quoted 43 lines · Uwe Kleine-König

From: Alex Williamson <[email protected]>

The vfio_ap and vfio_ccw drivers bind devices through driver_override. With the introduction of the support_driver_override flag in the core driver struct, it's now required for drivers to declare themselves as driver_override compatible, otherwise driver_override bindings will be denied or generate a kernel taint.

Set the support_driver_override flag on the vfio_ap and vfio_ccw driver structs.

---
 drivers/s390/cio/vfio_ccw_drv.c   | 1 +
 drivers/s390/crypto/vfio_ap_drv.c | 1 +
 2 files changed, 2 insertions(+)
drivers/s390/cio/vfio_ccw_drv.c +1 -0
index 7546bd137e4e..77765cde87f7 100644
@@ -375,6 +375,7 @@ static struct css_driver vfio_ccw_sch_driver = {
 	.drv = {
 		.name = "vfio_ccw",
 		.owner = THIS_MODULE,
+		.support_driver_override = true,
 	},
 	.subchannel_type = vfio_ccw_sch_ids,
 	.irq = vfio_ccw_sch_irq,
drivers/s390/crypto/vfio_ap_drv.c +1 -0
index 8e69ed286bb9..2d47dc643269 100644
@@ -62,6 +62,7 @@ static struct ap_driver vfio_ap_drv = {
 	.on_config_changed = vfio_ap_on_cfg_changed,
 	.on_scan_complete = vfio_ap_on_scan_complete,
 	.ids = ap_queue_ids,
+	.driver.support_driver_override = true,
 };
 
Quoted 5 lines

drivers/s390/crypto/vfio_ap_drv.c static struct ap_driver vfio_ap_drv = { · line 68

Quoted 1 line
drivers/s390/crypto/vfio_ap_drv.c +0 -0
 static void vfio_ap_matrix_dev_release(struct device *dev)

Hm, honestly I don' get this...
The driver_override is a bus thing. So the AP bus has support for driver_override.

Yes, and ap_device_probe() calls device_match_driver_override() which requires .driver.support_driver_override to be true to make the override work.

Quoted 8 lines

The sysfs functions are in drivers/s390/crypto/ap_queue.c. There exist currently
two drivers which shall be bound to an AP queue: vfio_ap and the default driver
which is the zcrypt driver "cex4queue". It sounds like the driver needs the tagging about driver_override support and then similar to the vfio_ap_drv.c the static struct ap_driver initialization may also need a .driver.support_driver_override = true line.

Can you please talk in form of a patch? I have difficulties following your words. (Hint: "the driver" is ambigous if you just talked about two different ones. Also what is "the static struct ap_driver initialization"?)

Alternatively something like: "I tested you changes and the xyz driver fails to bind, this is the respective kernel output: ..." would be welcome, too.

Best regards
Uwe

#10 lore raw

PATCH v2 5/10 uio: Allow untainted use of driver_override

In reply to Uwe Kleine-König

uio drivers support being matched to devices using the driver_override mechanism. Add .support_driver_override = true to continue having this mechanism work for them without tainting the kernel.

---
 drivers/uio/uio_dmem_genirq.c | 1 +
 drivers/uio/uio_hv_generic.c  | 1 +
 drivers/uio/uio_pci_generic.c | 1 +
 drivers/uio/uio_pdrv_genirq.c | 1 +
 4 files changed, 4 insertions(+)
drivers/uio/uio_dmem_genirq.c +1 -0
index 263c1ebc2d57..71992431770e 100644
@@ -309,6 +309,7 @@ static struct platform_driver uio_dmem_genirq = {
 	.driver = {
 		.name = DRIVER_NAME,
 		.of_match_table = of_match_ptr(uio_of_genirq_match),
+		.support_driver_override = true,
 	},
 };
 
drivers/uio/uio_hv_generic.c +1 -0
index dff92dcbb10d..255c2b5730f6 100644
@@ -407,6 +407,7 @@ static struct hv_driver hv_uio_drv = {
 	.id_table = hv_uio_id_table,
 	.probe = hv_uio_probe,
 	.remove = hv_uio_remove,
+	.driver.support_driver_override = true,
 };
 
 static int __init
drivers/uio/uio_pci_generic.c +1 -0
index c0cab3adb7b0..9785df9fc5de 100644
@@ -140,6 +140,7 @@ static struct pci_driver uio_pci_driver = {
 	.name = "uio_pci_generic",
 	.id_table = NULL, /* only dynamic id's */
 	.probe = probe,
+	.driver.support_driver_override = true,
 };
 
 module_pci_driver(uio_pci_driver);
drivers/uio/uio_pdrv_genirq.c +1 -0
index 87e43821ab0c..022178248962 100644
@@ -261,6 +261,7 @@ static struct platform_driver uio_pdrv_genirq = {
 	.driver = {
 		.name = DRIVER_NAME,
 		.of_match_table = of_match_ptr(uio_of_genirq_match),
+		.support_driver_override = true,
 	},
 };
 
Signature · 1 line
2.56.0.rc1
Reviewed-by Tested-by on v2 5/10
#11 lore raw

RE: [PATCH v2 05/10] uio: Allow untainted use of driver_override

From: Uwe Kleine-König <[email protected]> Sent: Friday, November 20, 2026 3:11 AM

Quoted 6 lines

uio drivers support being matched to devices using the driver_override mechanism. Add .support_driver_override = true to continue having this mechanism work for them without tainting the kernel.

Quoted 58 lines
---
 drivers/uio/uio_dmem_genirq.c | 1 +
 drivers/uio/uio_hv_generic.c  | 1 +
 drivers/uio/uio_pci_generic.c | 1 +
 drivers/uio/uio_pdrv_genirq.c | 1 +
 4 files changed, 4 insertions(+)
drivers/uio/uio_dmem_genirq.c +1 -0
index 263c1ebc2d57..71992431770e 100644
@@ -309,6 +309,7 @@ static struct platform_driver uio_dmem_genirq = {
 	.driver = {
 		.name = DRIVER_NAME,
 		.of_match_table = of_match_ptr(uio_of_genirq_match),
+		.support_driver_override = true,
 	},
 };
 
drivers/uio/uio_hv_generic.c +1 -0
index dff92dcbb10d..255c2b5730f6 100644
@@ -407,6 +407,7 @@ static struct hv_driver hv_uio_drv = {
 	.id_table = hv_uio_id_table,
 	.probe = hv_uio_probe,
 	.remove = hv_uio_remove,
+	.driver.support_driver_override = true,
 };
 
 static int __init
drivers/uio/uio_pci_generic.c +1 -0
index c0cab3adb7b0..9785df9fc5de 100644
@@ -140,6 +140,7 @@ static struct pci_driver uio_pci_driver = {
 	.name = "uio_pci_generic",
 	.id_table = NULL, /* only dynamic id's */
 	.probe = probe,
+	.driver.support_driver_override = true,
 };
 
 module_pci_driver(uio_pci_driver);
drivers/uio/uio_pdrv_genirq.c +1 -1
index 87e43821ab0c..022178248962 100644
@@ -261,6 +261,7 @@ static struct platform_driver uio_pdrv_genirq = {
 	.driver = {
 		.name = DRIVER_NAME,
 		.of_match_table = of_match_ptr(uio_of_genirq_match),
+		.support_driver_override = true,
 	},
 };
 
--

2.56.0.rc1

#12 lore raw

PATCH v2 6/10 spi: spidev: Allow untainted use of driver_override

In reply to Uwe Kleine-König

The spidev driver is designed to be bound to devices using the driver_override mechanism. Set .support_driver_override = true to continue having this mechanism work for it without tainting the kernel.

---
 drivers/spi/spidev.c | 1 +
 1 file changed, 1 insertion(+)
drivers/spi/spidev.c +1 -0
index 15d2aaeb08e9..0b50474ab20b 100644
@@ -843,6 +843,7 @@ static struct spi_driver spidev_spi_driver = {
 		.name =		"spidev",
 		.of_match_table = spidev_dt_ids,
 		.acpi_match_table = spidev_acpi_ids,
+		.support_driver_override = true,
 	},
 	.probe =	spidev_probe,
 	.remove =	spidev_remove,
Signature · 1 line
2.56.0.rc1
maintainer Acked-by on v2 6/10
#13 lore raw

Re: [PATCH v2 06/10] spi: spidev: Allow untainted use of driver_override

Quoted 3 lines · Uwe Kleine-König

The spidev driver is designed to be bound to devices using the driver_override mechanism. Set .support_driver_override = true to continue having this mechanism work for it without tainting the kernel.

#14 lore raw

PATCH v2 7/10 clk: imx: scu: Allow untainted use of driver_override

In reply to Uwe Kleine-König

The driver makes use of device_set_driver_override and so should be allowed to match via that driver_override. Note that the .probe() function assumes that platform_data is set for the device to bind, so it can very likely allow root to oops the kernel. This is not a regression however.

---
 drivers/clk/imx/clk-scu.c | 1 +
 1 file changed, 1 insertion(+)
drivers/clk/imx/clk-scu.c +1 -0
index 0d034e206c6b..7622890d1124 100644
@@ -626,6 +626,7 @@ static struct platform_driver imx_clk_scu_driver = {
 		.name = "imx-scu-clk",
 		.suppress_bind_attrs = true,
 		.pm = &imx_clk_scu_pm_ops,
+		.support_driver_override = true,
 	},
 	.probe = imx_clk_scu_probe,
 };
Signature · 1 line
2.56.0.rc1
#15 lore raw

PATCH v2 8/10 remoteproc: Allow untainted use of driver_override

In reply to Uwe Kleine-König

The two drivers are expected to be bound to devices using a driver override. (See the respective use of device_set_driver_override()) and so should be allowed to match via that mechanism.

---
 drivers/rpmsg/rpmsg_ctrl.c | 1 +
 drivers/rpmsg/rpmsg_ns.c   | 5 ++++-
 2 files changed, 5 insertions(+), 1 deletion(-)
drivers/rpmsg/rpmsg_ctrl.c +1 -0
index 5c19188d0294..79ae495c9c07 100644
@@ -210,6 +210,7 @@ static struct rpmsg_driver rpmsg_ctrldev_driver = {
 	.remove = rpmsg_ctrldev_remove,
 	.drv = {
 		.name = "rpmsg_ctrl",
+		.support_driver_override = true,
 	},
 };
 
drivers/rpmsg/rpmsg_ns.c +4 -1
index bde8c8d433e0..820dabbc9500 100644
@@ -96,7 +96,10 @@ static int rpmsg_ns_probe(struct rpmsg_device *rpdev)
 }
 
 static struct rpmsg_driver rpmsg_ns_driver = {
-	.drv.name = KBUILD_MODNAME,
+	.drv = {
+		.name = KBUILD_MODNAME,
+		.support_driver_override = true,
+	},
 	.probe = rpmsg_ns_probe,
 };
 
Signature · 1 line
2.56.0.rc1
maintainer Reviewed-by on v2 8/10
#16 lore raw

Re: [PATCH v2 08/10] remoteproc: Allow untainted use of driver_override

Quoted 11 lines · Uwe Kleine-König

The two drivers are expected to be bound to devices using a driver override. (See the respective use of device_set_driver_override()) and so should be allowed to match via that mechanism.

---
 drivers/rpmsg/rpmsg_ctrl.c | 1 +
 drivers/rpmsg/rpmsg_ns.c   | 5 ++++-
 2 files changed, 5 insertions(+), 1 deletion(-)
Quoted 31 lines
drivers/rpmsg/rpmsg_ctrl.c +1 -0
index 5c19188d0294..79ae495c9c07 100644
@@ -210,6 +210,7 @@ static struct rpmsg_driver rpmsg_ctrldev_driver = {
        .remove = rpmsg_ctrldev_remove,
        .drv = {
                .name = "rpmsg_ctrl",
+               .support_driver_override = true,
        },
 };
 
drivers/rpmsg/rpmsg_ns.c +4 -2
index bde8c8d433e0..820dabbc9500 100644
@@ -96,7 +96,10 @@ static int rpmsg_ns_probe(struct rpmsg_device *rpdev)
 }
 
 static struct rpmsg_driver rpmsg_ns_driver = {
-       .drv.name = KBUILD_MODNAME,
+       .drv = {
+               .name = KBUILD_MODNAME,
+               .support_driver_override = true,
+       },
        .probe = rpmsg_ns_probe,
 };
 
--

2.56.0.rc1

#17 lore raw

PATCH v2 9/10 slimbus: qcom-ngd-ctrl: Allow untainted use of driver_override

In reply to Uwe Kleine-König

The driver is expected to be bound to devices using a driver override---see the respective use of device_set_driver_override()---and so should be allowed to match via that mechanism.

---
 drivers/slimbus/qcom-ngd-ctrl.c | 1 +
 1 file changed, 1 insertion(+)
drivers/slimbus/qcom-ngd-ctrl.c +1 -0
index 8da75c4b50e5..f501740fa4a6 100644
@@ -1837,6 +1837,7 @@ static struct platform_driver qcom_slim_ngd_driver = {
 	.driver	= {
 		.name = QCOM_SLIM_NGD_DRV_NAME,
 		.pm = &qcom_slim_ngd_dev_pm_ops,
+		.support_driver_override = true,
 	},
 };
 
Signature · 1 line
2.56.0.rc1
#18 lore raw

PATCH v2 10/10 ASoC: samsung: i2s: Allow untainted use of driver_override

In reply to Uwe Kleine-König

The driver is expected to be bound to devices using a driver override---see the respective use of device_set_driver_override()---and so should be allowed to match via that mechanism.

---
 sound/soc/samsung/i2s.c | 1 +
 1 file changed, 1 insertion(+)
sound/soc/samsung/i2s.c +1 -0
index 8a938d0d79f9..fdddd3d4eb65 100644
@@ -1740,6 +1740,7 @@ static struct platform_driver samsung_i2s_driver = {
 		.name = "samsung-i2s",
 		.of_match_table = of_match_ptr(exynos_i2s_match),
 		.pm = pm_ptr(&samsung_i2s_pm),
+		.support_driver_override = true,
 	},
 };
 
Signature · 1 line
2.56.0.rc1
maintainer
#19 lore raw
In reply to Uwe Kleine-König
Quoted 8 lines · Uwe Kleine-König

It is sensible to merge the driver core change (patch #1) together with the individual driver patches setting .support_driver_override because with the driver core change the driver override doesn't work any more for the unmodified drivers (unless the new kernel parameter is provided and a taint is accepted). So I suggest (once there is agreement that the patches are fine) to merge them all in one go and provide a stable branch in case a driver was missed to be adapted. I can create such a branch and ask Greg or Linus to pull it, or Greg can do that.

I think it should be good enough to just take them all through driver-core. If it turns out in linux-next that we missed a driver we can always add another patch to the driver-core tree.

Those patches only set support_driver_override = true, so there shouldn't be interactions with other trees.

So, let's keep it simple. I can pick this up once we have all the ACKs.

Thanks,
Danilo