Re: [PATCH v4 0/3] serdev: handle devices without an ACPI or device tree node

[An Wu]

Hi Makus
    According to sashiko:
'''
In drivers/tty/serdev/core.c, serdev_controller_alloc() places controller
devices on serdev_bus_type:

    device_initialize(&ctrl->dev);
    ctrl->dev.type = &serdev_ctrl_type;
    ctrl->dev.bus = &serdev_bus_type;

When yt2_1380_fc_module_init() calls serdev_device_driver_register() for
yt2_1380_fc_serdev_driver (or when a new serdev_controller is added), the
driver core runs serdev_device_match() against every device on
serdev_bus_type, including struct serdev_controller instances:

static int serdev_device_match(struct device *dev, const struct
device_driver *drv)
{
        const struct serdev_device *serdev = to_serdev_device(dev);
        const struct serdev_device_driver *sdrv = to_serdev_device_driver(drv);

        if (acpi_driver_match_device(dev, drv))
                return 1;

        if (of_driver_match_device(dev, drv))
                return 1;

        return !!serdev_match_id(sdrv->id_table, serdev->modalias);
}

Because serdev_device_match() does not check is_serdev_device(dev) before
casting dev with to_serdev_device(dev), and yt2_1380_fc_serdev_driver has no
ACPI or OF match table, serdev_device_match() falls through to
serdev_match_id(sdrv->id_table, serdev->modalias) on struct serdev_controller
devices.

In struct serdev_device, modalias[SERDEV_NAME_SIZE] starts at offset
sizeof(struct device) + 4, which in struct serdev_controller overlaps the
struct device *host pointer, ctrl->nr, ctrl->serdev, ctrl->ops, and extends 12
bytes past sizeof(struct serdev_controller). When the overlapping byte at
modalias[0] is non-zero, serdev_match_id() passes that non-string memory and
trailing out-of-bounds bytes to strcmp().

Should serdev_device_match() verify is_serdev_device(dev) before casting dev
to struct serdev_device?
'''

If we remove is_serdev_device could cause some problem due to the
serdev_controller could also
run into this function and "serdev_match_id(sdrv->id_table,
serdev->modalias);" can be a potential risk.

I will add the "is_serdev_device" back and send v5 out.

BR
    An


On Sat, Oct 24, 2026 at 10:29 AM ChunAn Wu <an.wu@xxxxxxxxxxxxx> wrote:
>
> Driver-created serdev clients without an ACPI companion or device tree
> node currently have no modalias and cannot be matched to a driver:
> their modalias sysfs read returns -ENODEV and their creators force a
> bind with device_driver_attach().
>
> Backlight function on dell-uart-backlight is not affected: that class
> device is registered independently and gets its own add uevent. The
> user-visible breakage is the failing modalias read on an existing
> serdev child -- Checkbox dkms_info_attachment aborts with
> "OSError: [Errno 19] No such device".
>
> Resolve the two long-standing serdev-core TODOs by letting the creator
> pass a modalias to serdev_device_add() and giving the serdev driver an
> id_table, modelled on the SPI bus. ACPI and OF match first and are
> unchanged. Convert dell-uart-backlight and the Lenovo Yoga Tab 2 Pro
> fastcharger to this path and drop their force-attach workarounds.
>
> Changes in v4:
> - Patch 1: drop the dead is_serdev_device() guard from the match
>   callback, and dereference serdev once at the top
> - Patch 1: drop the "large enough" comment on SERDEV_NAME_SIZE
> - Patch 1: in struct serdev_device_driver, place id_table right under
>   driver, next to driver->of_match_table and driver->acpi_match_table
> - Patch 1: merge serdev_device_set_modalias() into serdev_device_add()
>   so the modalias is set at the only point where forgetting it would
>   matter; internal firmware callers pass NULL
> - All three patches: tightened commit messages and moved the v2/v3
>   changelog narrative out of Patch 1 into this cover letter
>
> Changes in v3:
> - Rename struct serdev_device::name to ::modalias; dev_name() is
>   unchanged
> - Model matching on the SPI bus: add struct serdev_device_id and
>   id_table; drop driver->name matching
> - Add serdev_device_set_modalias() helper and SERDEV_MODULE_PREFIX
> - No MODULE_DEVICE_TABLE(serdev, ...) support: both in-tree
>   firmware-less clients are created by the same module that provides
>   their driver, so udev has nothing to autoload
>
> Changes in v2:
> - Instead of silently returning success with no modalias, use a
>   creator-supplied name as a "serial:<name>" modalias fallback and as
>   the matching key, so clients can bind through normal serdev matching
>
> v3: https://lore.kernel.org/lkml/20261023052717.2577234-1-an.wu@xxxxxxxxxxxxx/
> v2: https://lore.kernel.org/lkml/20261022032842.2492572-1-an.wu@xxxxxxxxxxxxx/
> v1: https://lore.kernel.org/lkml/20261021092124.2306053-1-an.wu@xxxxxxxxxxxxx/
>
> ChunAn Wu (3):
>   serdev: handle devices without an ACPI or device tree node
>   platform/x86: dell-uart-backlight: Bind through normal serdev matching
>   platform/x86: lenovo-yoga-tab2-pro-1380-fastcharger: Bind through
>     normal serdev matching
>
>  MAINTAINERS                                   |  1 +
>  .../platform/x86/dell/dell-uart-backlight.c   | 33 ++++------
>  .../lenovo/yoga-tab2-pro-1380-fastcharger.c   | 25 +++----
>  .../platform/x86/x86-android-tablets/core.c   |  2 +-
>  drivers/power/sequencing/pwrseq-pcie-m2.c     |  2 +-
>  drivers/tty/serdev/core.c                     | 65 ++++++++++++++-----
>  include/linux/device-id/serdev.h              | 19 ++++++
>  include/linux/serdev.h                        | 12 +++-
>  8 files changed, 105 insertions(+), 54 deletions(-)
>  create mode 100644 include/linux/device-id/serdev.h
>
>
> base-commit: 9e947d96915af63e28896caae85a81b74855621b
> --
> 2.34.1
>