tools/memory-model: Make judgelitmus.sh handle hardware verifications
authorPaul E. McKenney <paulmck@kernel.org>
Tue, 19 Mar 2019 21:39:10 +0000 (14:39 -0700)
committerPaul E. McKenney <paulmck@kernel.org>
Wed, 1 Dec 2021 01:48:14 +0000 (17:48 -0800)
commit2ced87bb0e49dc288b053ed9a0d69e8f3aa6c5bb
treed63968d24c680bf23f7e3d921e261d3b1944c830
parent414c624908281d0c753a1eef95c343781ee50700
tools/memory-model: Make judgelitmus.sh handle hardware verifications

This commit makes the judgelitmus.sh script check the --hw argument
(AKA the LKMM_HW_MAP_FILE environment variable) and to adjust its
judgment for a run where a C-language litmus test has been translated to
assembly and the assembly version verified.  In this case, the assembly
verification output is checked against the C-language script's "Result:"
comment.  However, because hardware can be stronger than LKMM requires,
the judgelitmus.sh script forgives verification mismatches featuring
a "Sometimes" in the C-language script and an "Always" or "Never"
assembly-language verification.

Note that deadlock is not forgiven, however, this should not normally be
an issue given that C-language tests containing locking, RCU, or SRCU
cannot be translated to assembly.  However, this issue can crop up in
litmus tests that mimic deadlock by using the "filter" clause to ignore
all executions.  It can also crop up when certain herd arguments are
used to autofilter everything that does not match the "exists" clause
in cases where the "exists" clause cannot be satisfied.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
tools/memory-model/scripts/README
tools/memory-model/scripts/judgelitmus.sh