Re: [PATCH v2 09/15] peci: Add sysfs interface for PECI bus

[Dan Williams <dan.j.williams@intel.com>]

On Tue, Aug 3, 2021 at 4:35 AM Iwona Winiarska
<iwona.winiarska@intel.com> wrote:
>
> PECI devices may not be discoverable at the time when PECI controller is
> being added (e.g. BMC can boot up when the Host system is still in S5).
> Since we currently don't have the capabilities to figure out the Host
> system state inside the PECI subsystem itself, we have to rely on
> userspace to do it for us.
>
> In the future, PECI subsystem may be expanded with mechanisms that allow
> us to avoid depending on userspace interaction (e.g. CPU presence could
> be detected using GPIO, and the information on whether it's discoverable
> could be obtained over IPMI).

Thanks for this detail.

> Unfortunately, those methods may ultimately not be available (support
> will vary from platform to platform), which means that we still need
> platform independent method triggered by userspace.
>
> Signed-off-by: Iwona Winiarska <iwona.winiarska@intel.com>
> ---
>  Documentation/ABI/testing/sysfs-bus-peci | 16 +++++
>  drivers/peci/Makefile                    |  2 +-
>  drivers/peci/core.c                      |  3 +-
>  drivers/peci/device.c                    |  1 +
>  drivers/peci/internal.h                  |  5 ++
>  drivers/peci/sysfs.c                     | 82 ++++++++++++++++++++++++
>  6 files changed, 107 insertions(+), 2 deletions(-)
>  create mode 100644 Documentation/ABI/testing/sysfs-bus-peci
>  create mode 100644 drivers/peci/sysfs.c
>
> diff --git a/Documentation/ABI/testing/sysfs-bus-peci b/Documentation/ABI/testing/sysfs-bus-peci
> new file mode 100644
> index 000000000000..56c2b2216bbd
> --- /dev/null
> +++ b/Documentation/ABI/testing/sysfs-bus-peci
> @@ -0,0 +1,16 @@
> +What:          /sys/bus/peci/rescan
> +Date:          July 2021
> +KernelVersion: 5.15
> +Contact:       Iwona Winiarska <iwona.winiarska@intel.com>
> +Description:
> +               Writing a non-zero value to this attribute will
> +               initiate scan for PECI devices on all PECI controllers
> +               in the system.
> +
> +What:          /sys/bus/peci/devices/<controller_id>-<device_addr>/remove
> +Date:          July 2021
> +KernelVersion: 5.15
> +Contact:       Iwona Winiarska <iwona.winiarska@intel.com>
> +Description:
> +               Writing a non-zero value to this attribute will
> +               remove the PECI device and any of its children.
> diff --git a/drivers/peci/Makefile b/drivers/peci/Makefile
> index c5f9d3fe21bb..917f689e147a 100644
> --- a/drivers/peci/Makefile
> +++ b/drivers/peci/Makefile
> @@ -1,7 +1,7 @@
>  # SPDX-License-Identifier: GPL-2.0-only
>
>  # Core functionality
> -peci-y := core.o request.o device.o
> +peci-y := core.o request.o device.o sysfs.o
>  obj-$(CONFIG_PECI) += peci.o
>
>  # Hardware specific bus drivers
> diff --git a/drivers/peci/core.c b/drivers/peci/core.c
> index d143f1a7fe98..c473acb3c2a0 100644
> --- a/drivers/peci/core.c
> +++ b/drivers/peci/core.c
> @@ -34,7 +34,7 @@ struct device_type peci_controller_type = {
>         .release        = peci_controller_dev_release,
>  };
>
> -static int peci_controller_scan_devices(struct peci_controller *controller)
> +int peci_controller_scan_devices(struct peci_controller *controller)
>  {
>         int ret;
>         u8 addr;
> @@ -159,6 +159,7 @@ EXPORT_SYMBOL_NS_GPL(devm_peci_controller_add, PECI);
>
>  struct bus_type peci_bus_type = {
>         .name           = "peci",
> +       .bus_groups     = peci_bus_groups,
>  };
>
>  static int __init peci_init(void)
> diff --git a/drivers/peci/device.c b/drivers/peci/device.c
> index 32811248997b..d77d9dabd51e 100644
> --- a/drivers/peci/device.c
> +++ b/drivers/peci/device.c
> @@ -110,5 +110,6 @@ static void peci_device_release(struct device *dev)
>  }
>
>  struct device_type peci_device_type = {
> +       .groups         = peci_device_groups,
>         .release        = peci_device_release,
>  };
> diff --git a/drivers/peci/internal.h b/drivers/peci/internal.h
> index 57d11a902c5d..978e12c8e1d3 100644
> --- a/drivers/peci/internal.h
> +++ b/drivers/peci/internal.h
> @@ -8,6 +8,7 @@
>  #include <linux/types.h>
>
>  struct peci_controller;
> +struct attribute_group;
>  struct peci_device;
>  struct peci_request;
>
> @@ -19,12 +20,16 @@ struct peci_request *peci_request_alloc(struct peci_device *device, u8 tx_len, u
>  void peci_request_free(struct peci_request *req);
>
>  extern struct device_type peci_device_type;
> +extern const struct attribute_group *peci_device_groups[];
>
>  int peci_device_create(struct peci_controller *controller, u8 addr);
>  void peci_device_destroy(struct peci_device *device);
>
>  extern struct bus_type peci_bus_type;
> +extern const struct attribute_group *peci_bus_groups[];

To me, sysfs.c is small enough to just fold into core.c, then no need
to declare public attribute arrays like this, but up to you if you
prefer the sysfs.c split.

>
>  extern struct device_type peci_controller_type;
>
> +int peci_controller_scan_devices(struct peci_controller *controller);
> +
>  #endif /* __PECI_INTERNAL_H */
> diff --git a/drivers/peci/sysfs.c b/drivers/peci/sysfs.c
> new file mode 100644
> index 000000000000..db9ef05776e3
> --- /dev/null
> +++ b/drivers/peci/sysfs.c
> @@ -0,0 +1,82 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +// Copyright (c) 2021 Intel Corporation
> +
> +#include <linux/device.h>
> +#include <linux/kernel.h>
> +#include <linux/peci.h>
> +
> +#include "internal.h"
> +
> +static int rescan_controller(struct device *dev, void *data)
> +{
> +       if (dev->type != &peci_controller_type)
> +               return 0;
> +
> +       return peci_controller_scan_devices(to_peci_controller(dev));
> +}
> +
> +static ssize_t rescan_store(struct bus_type *bus, const char *buf, size_t count)
> +{
> +       bool res;
> +       int ret;
> +
> +       ret = kstrtobool(buf, &res);
> +       if (ret)
> +               return ret;
> +
> +       if (!res)
> +               return count;
> +
> +       ret = bus_for_each_dev(&peci_bus_type, NULL, NULL, rescan_controller);
> +       if (ret)
> +               return ret;
> +
> +       return count;
> +}
> +static BUS_ATTR_WO(rescan);
> +
> +static struct attribute *peci_bus_attrs[] = {
> +       &bus_attr_rescan.attr,
> +       NULL
> +};
> +
> +static const struct attribute_group peci_bus_group = {
> +       .attrs = peci_bus_attrs,
> +};
> +
> +const struct attribute_group *peci_bus_groups[] = {
> +       &peci_bus_group,
> +       NULL
> +};
> +
> +static ssize_t remove_store(struct device *dev, struct device_attribute *attr,
> +                           const char *buf, size_t count)
> +{
> +       struct peci_device *device = to_peci_device(dev);
> +       bool res;
> +       int ret;
> +
> +       ret = kstrtobool(buf, &res);
> +       if (ret)
> +               return ret;
> +
> +       if (res && device_remove_file_self(dev, attr))
> +               peci_device_destroy(device);

How do you solve races between sysfs device remove and controller
device remove? Looks like double-free at first glance. Have a look at
the  kill_device() helper as one way to resolve this double-delete
race..

> +
> +       return count;
> +}
> +static DEVICE_ATTR_IGNORE_LOCKDEP(remove, 0200, NULL, remove_store);
> +
> +static struct attribute *peci_device_attrs[] = {
> +       &dev_attr_remove.attr,
> +       NULL
> +};
> +
> +static const struct attribute_group peci_device_group = {
> +       .attrs = peci_device_attrs,
> +};
> +
> +const struct attribute_group *peci_device_groups[] = {
> +       &peci_device_group,
> +       NULL
> +};
> --
> 2.31.1
>