On Fri, 11 Jan 2019 at 20:39, Bhupesh Sharma bhsharma@redhat.com wrote:
Hi Sumit,
Thanks for the patch. Some nitpicks in-line:
Apologies for delay in my response as I was on leave. Please find my comments inline.
Will include these nitpicks in v4 but let me wait for any further major comments.
On 01/11/2019 05:17 PM, Sumit Garg wrote:
Introduce a generic TEE bus driver concept for TEE based kernel drivers which would like to communicate with TEE based devices/services.
In this TEE bus concept, devices/services are identified via Universally Unique Identifier (UUID) and drivers register a table of device UUIDs which they can support.
So this TEE bus framework registers a match() callback function which iterates over the driver UUID table to find a corresponding match for device UUID. If a match is found, then this particular device is probed via corresponding probe api registered by the driver. This process happens whenever a device or a driver is registered with TEE bus.
Also this framework allows for device enumeration to be specific to corresponding TEE implementation like OP-TEE etc.
Signed-off-by: Sumit Garg sumit.garg@linaro.org
drivers/tee/tee_core.c | 43 ++++++++++++++++++++++++++++++++++++++++--- include/linux/tee_drv.h | 36 ++++++++++++++++++++++++++++++++++++ 2 files changed, 76 insertions(+), 3 deletions(-)
diff --git a/drivers/tee/tee_core.c b/drivers/tee/tee_core.c index 7b2bb4c..a685940 100644 --- a/drivers/tee/tee_core.c +++ b/drivers/tee/tee_core.c @@ -15,7 +15,6 @@ #define pr_fmt(fmt) "%s: " fmt, __func__
#include <linux/cdev.h> -#include <linux/device.h> #include <linux/fs.h> #include <linux/idr.h> #include <linux/module.h> @@ -1027,6 +1026,30 @@ int tee_client_invoke_func(struct tee_context *ctx, } EXPORT_SYMBOL_GPL(tee_client_invoke_func);
+static int tee_client_device_match(struct device *dev,
struct device_driver *drv)+{
struct tee_client_device *tee_device;const struct tee_client_device_id *id_table;tee_device = to_tee_client_device(dev);id_table = to_tee_client_driver(drv)->id_table;while (!uuid_is_null(&id_table->uuid)) {if (uuid_equal(&tee_device->id.uuid, &id_table->uuid))return 1;id_table++;}return 0;+}
+struct bus_type tee_bus_type = {
.name = "tee",.match = tee_client_device_match,+}; +EXPORT_SYMBOL_GPL(tee_bus_type);
- static int __init tee_init(void) { int rc;
@@ -1040,10 +1063,23 @@ static int __init tee_init(void) rc = alloc_chrdev_region(&tee_devt, 0, TEE_NUM_DEVICES, "tee"); if (rc) { pr_err("failed to allocate char dev region\n");
class_destroy(tee_class);tee_class = NULL;
goto chrdev_err; }rc = bus_register(&tee_bus_type);if (rc) {pr_err("failed to register tee bus\n");goto bus_err;}return 0;+bus_err:
unregister_chrdev_region(tee_devt, TEE_NUM_DEVICES);+chrdev_err:
class_destroy(tee_class);tee_class = NULL;Hmm.. these error paths/labels look out-of-order. See 'drivers/i2c/i2c-dev.c' for example.
Normally our error paths in an __init function is of the same order as the __exit function implementation. So this should be changed to something like:
+out_unreg_class:
- class_destroy(tee_class);
- tee_class = NULL;
+out_unreg_chrdev:
- unregister_chrdev_region(tee_devt, TEE_NUM_DEVICES);
It seems that prior to this patch also allocation order in __init function differed from release order in __exit function. Will correct that too along with this.
return rc;}
@@ -1052,6 +1088,7 @@ static void __exit tee_exit(void) class_destroy(tee_class); tee_class = NULL; unregister_chrdev_region(tee_devt, TEE_NUM_DEVICES);
bus_unregister(&tee_bus_type);Since the __exit function order is the reverse of the __init one, lets reorder this as:
bus_unregister(&tee_bus_type); class_destroy(tee_class); tee_class = NULL; unregister_chrdev_region(tee_devt, TEE_NUM_DEVICES);
Ok.
See 'drivers/i2c/i2c-dev.c' for example.
}
subsys_initcall(tee_init); diff --git a/include/linux/tee_drv.h b/include/linux/tee_drv.h index 6cfe058..ed16bf1 100644 --- a/include/linux/tee_drv.h +++ b/include/linux/tee_drv.h @@ -20,6 +20,8 @@ #include <linux/kref.h> #include <linux/list.h> #include <linux/tee.h> +#include <linux/device.h> +#include <linux/uuid.h>
If possible, let's keep alphabetical order of header files when making changes in the patch.
Ok.
/*
- The file describes the API provided by the generic TEE driver to the
@@ -538,4 +540,38 @@ static inline bool tee_param_is_memref(struct tee_param *param) } }
+extern struct bus_type tee_bus_type;
+/**
- struct tee_client_device_id - tee based device identifier
- @uuid: For TEE based client devices we use the device uuid
as the identifier.- */
+struct tee_client_device_id {
uuid_t uuid;+};
Hmm.. Do we really need a struct for a single element, rather lets use a simple typedef here.
Agree, will rather use typedef here.
+/**
- struct tee_client_device - tee based device
- @id: device identifier
- @dev: device structure
- */
+struct tee_client_device {
struct tee_client_device_id id;struct device dev;+};
Add a blank line here.
Ok.
+#define to_tee_client_device(d) container_of(d, struct tee_client_device, dev)
+/**
- struct tee_client_driver - tee client driver
- @id_table: device id table supported by this driver
- @driver: driver structure
- */
+struct tee_client_driver {
const struct tee_client_device_id *id_table;struct device_driver driver;+};
Add a blank line here.
Ok.
-Sumit
+#define to_tee_client_driver(d) \
container_of(d, struct tee_client_driver, driver)- #endif /*__TEE_DRV_H*/
Thanks, Bhupesh