I noticed 4.19.35 got a backport of mainline 4abb951b, but it appears to be a duplicate backport that landed in the wrong function. We can see this in the stable-queue repo:
stable-queue$ find . -name '*acpica-aml-interpreter-add-region-addr*' |grep 4.19 ./releases/4.19.6/acpica-aml-interpreter-add-region-addresses-in-global-list-during-initialization.patch ./releases/4.19.3/revert-acpica-aml-interpreter-add-region-addresses-in.patch ./releases/4.19.35/acpica-aml-interpreter-add-region-addresses-in-global-list-during-initialization.patch ./releases/4.19.2/acpica-aml-interpreter-add-region-addresses-in-global-list-during-initialization.patch
So it was added to 4.19.2, reverted in .3, re-added in .6, and then finally patched into a similar looking but wrong function in .35
If we diff the .6 and .35 versions, we see the function difference:
-@@ -417,6 +417,10 @@ acpi_ds_eval_region_operands(struct acpi +@@ -523,6 +523,10 @@ acpi_ds_eval_table_region_operands(struc
I don't know what the history is/was around the 2/3/6 churn, but the re-addition in 4.19.35 to a different function sure looks wrong.
The commit adds a call "status = acpi_ut_add_address_range(..." and if we check mainline, there is only one in that file, but in 4.19.35+ there now are two calls - since the two functions had similar context and comments, it isn't hard to see how patch could/would apply it a 2nd time in the wrong place.
I didn't check if any of the other currently maintained linux-stable versions also had this possible issue.
Paul.