Changes in v14 (addressing Ilpo Järvinen's v13 review of patch 2/2):

  - Restore #include <linux/compiler_attributes.h> and <linux/spinlock.h>,
    which were accidentally absent from v13.
  - Restore spinlock_t lock in struct yb9_kbdock_data, spin_lock_init(),
    and spin_lock/spin_unlock around input_report_switch + input_sync
    in yb9_kbdock_report().
  - Restore pointer-out yb9_kbdock_query(d, u32 *bkbd) and removal of
    zero-initializer from struct wmi_buffer out.
  - Restore int return from yb9_kbdock_sync() and error propagation in
    yb9_kbdock_resume() and yb9_kbdock_block_probe().

Changes in v13:

  - Resend of v12: v12 was corrupted in transit.  Root cause: the ABI
    documentation uses RST table separator lines (==  ====...) which
    contain '=' characters; git-send-email selected quoted-printable
    transfer encoding due to non-ASCII in the changelog (UTF-8 author
    name), encoding '=' as '=3D' and breaking git-am on the receiving
    end.  Fixed by setting sendemail.transferEncoding=8bit.  No content
    changes from v12.

Changes in v12 (addressing Ilpo Järvinen's v11 review of patch 2/2):

  - Add #include <linux/sysfs.h> (for sysfs_emit)
  - Add #include <linux/types.h> (for __le32 and __packed)

Changes in v11 (addressing Armin Wolf's v10 review of patch 2/2):

  - ABI doc: correct Date to "June 2027" and KernelVersion to "7.3";
    these fields should reflect when the ABI lands in the kernel, not
    the submission date.

Patch 1/2 (lenovo-ymc) is unchanged from v10.

Changes in v10 (addressing Armin Wolf's v9 review of patch 2/2):

  - Update ABI doc: Date: June 2026, KernelVersion: 6.16
  - Add #include <linux/compiler_attributes.h> and <linux/types.h>
  - yb9_kbdock_query(): change return to pointer-out u32 *bkbd to avoid
    a 0xFFFFFFFF status value being misread as -EPERM
  - struct yb9_kbdock_data: add spinlock_t lock
  - yb9_kbdock_report(): hold lock across input_report_switch + input_sync
  - yb9_kbdock_sync(): return int, propagate query error to callers
  - yb9_kbdock_resume(), yb9_kbdock_block_probe(): propagate sync error
  - struct wmi_buffer out: remove zero-initializer (not needed)
  - yb9_kbdock_notify_new(): drop manual length check; use
    .min_event_size = sizeof(__le32) in struct wmi_driver instead

Patch 1/2 (lenovo-ymc) is unchanged from v9.

Dave Carey (2):
  platform/x86/lenovo: lenovo-ymc: Suppress probe on Yoga Book 9 14IAH10
  platform/x86/lenovo: Add Yoga Book 9 keyboard dock detection driver

 Documentation/ABI/testing/sysfs-driver-lenovo-yb9-kbdock |  19 ++
 MAINTAINERS                                               |   7 +
 drivers/platform/x86/lenovo/Kconfig                       |  14 +
 drivers/platform/x86/lenovo/Makefile                      |   1 +
 drivers/platform/x86/lenovo/yb9-kbdock.c                  | 324 ++++++++++++
 drivers/platform/x86/lenovo/ymc.c                         |  19 ++
 6 files changed, 384 insertions(+)

2.55.0

On Wed, 22 Jul 2026, Dave Carey wrote:

> Changes in v14 (addressing Ilpo Järvinen's v13 review of patch 2/2):
> 
>   - Restore #include <linux/compiler_attributes.h> and <linux/spinlock.h>,
>     which were accidentally absent from v13.
>   - Restore spinlock_t lock in struct yb9_kbdock_data, spin_lock_init(),
>     and spin_lock/spin_unlock around input_report_switch + input_sync
>     in yb9_kbdock_report().
>   - Restore pointer-out yb9_kbdock_query(d, u32 *bkbd) and removal of
>     zero-initializer from struct wmi_buffer out.
>   - Restore int return from yb9_kbdock_sync() and error propagation in
>     yb9_kbdock_resume() and yb9_kbdock_block_probe().

Hi,

Unfortunately, this still seems to miss one non-cosmetic change (diff of 
diffs, v12 vs v14):

diff --git a/drivers/platform/x86/lenovo/yb9-kbdock.c b/drivers/platform/x86/lenovo/yb9>
 new file mode 100644
-index 0000000..0000000
+index 0000000..40a2566
 --- /dev/null
 +++ b/drivers/platform/x86/lenovo/yb9-kbdock.c
-@@ -0,0 +1,322 @@
+@@ -0,0 +1,324 @@
 +// SPDX-License-Identifier: GPL-2.0-or-later
 +/*
 + * Lenovo Yoga Book 9 keyboard-dock detection
@@ -421,8 +449,10 @@
 +      /*
 +       * _WED(0xEB) returns EC.BKBD directly as a 32-bit integer
 +       * (LENOVO_BTKBD_EVENT WmiDataId(1) uint32 Status).
-+       * Short-buffer guard is handled by .min_event_size below.
 +       */
++      if (!data || data->length < sizeof(__le32))
++              return;
++
 +      u32 bkbd = le32_to_cpu(*(const __le32 *)data->data);
 +
 +      blocking_notifier_call_chain(&yb9_kbdock_chain_head, bkbd, NULL);
@@ -441,7 +471,6 @@
 +      .id_table       = yb9_kbdock_event_id_table,
 +      .no_singleton   = true,
 +      .notify_new     = yb9_kbdock_notify_new,
-+      .min_event_size = sizeof(__le32),
 +};
 +
 +/* ------------------------------------------------------------------


> Changes in v13:

[...snip...]

> Changes in v10 (addressing Armin Wolf's v9 review of patch 2/2):

[...snip...]

>   - yb9_kbdock_notify_new(): drop manual length check; use
>     .min_event_size = sizeof(__le32) in struct wmi_driver instead
> 
> Patch 1/2 (lenovo-ymc) is unchanged from v9.


...And there's no need to apologize. :-)


I personally keep earlier versions of the patches I've sent myself around 
as files so I can easily do diff-of-diffs against the previous version. It 
has saved me sooo many time from making fool of myself.

I actually normally write the patch version history bullets based on that 
diff so I won't forget anything important from it. Nor hallucinate some 
changes that I thought I'd made but didn't (=> back to the drawing board).


If you need the old version you can probably get them from lore archive if 
you don't have them around anymore (but by not placing them cleanly into 
own threads one needs to be careful to get the right ones).

-- 
 i.