Documentation/x86: Update split lock documentation
authorTony Luck <tony.luck@intel.com>
Wed, 15 Mar 2023 22:57:22 +0000 (15:57 -0700)
committerJonathan Corbet <corbet@lwn.net>
Thu, 23 Mar 2023 18:02:35 +0000 (12:02 -0600)
commit b041b525dab9 ("x86/split_lock: Make life miserable for split
lockers") added a delay and serialization of split locks. Commit
727209376f49 ("x86/split_lock: Add sysctl to control the misery mode")
provided a sysctl to turn off the misery.

Update the split lock documentation to describe the current state of
the code.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/20230315225722.104607-1-tony.luck@intel.com
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Documentation/x86/buslock.rst

index 7c051e714943cb7bc625b332dd36669e13dbb839..31ec0ef780868304f2dc44616fbeeb9436ac042e 100644 (file)
@@ -53,8 +53,14 @@ parameter "split_lock_detect". Here is a summary of different options:
 |off              |Do nothing                  |Do nothing             |
 +------------------+----------------------------+-----------------------+
 |warn             |Kernel OOPs                 |Warn once per task and |
-|(default)        |Warn once per task and      |and continues to run.  |
-|                 |disable future checking     |                       |
+|(default)        |Warn once per task, add a   |and continues to run.  |
+|                 |delay, add synchronization  |                       |
+|                 |to prevent more than one    |                       |
+|                 |core from executing a       |                       |
+|                 |split lock in parallel.     |                       |
+|                 |sysctl split_lock_mitigate  |                       |
+|                 |can be used to avoid the    |                       |
+|                 |delay and synchronization   |                       |
 |                 |When both features are      |                       |
 |                 |supported, warn in #AC      |                       |
 +------------------+----------------------------+-----------------------+