From: Yuyang Du <duyuyang@gmail.com>
To: peterz@infradead.org, will.deacon@arm.com, mingo@kernel.org
Cc: bvanassche@acm.org, ming.lei@redhat.com, frederic@kernel.org,
tglx@linutronix.de, linux-kernel@vger.kernel.org,
longman@redhat.com, paulmck@linux.vnet.ibm.com,
boqun.feng@gmail.com, Yuyang Du <duyuyang@gmail.com>
Subject: [PATCH v4 08/30] locking/lockdep: Skip checks if direct dependency is already present
Date: Thu, 29 Aug 2019 16:31:10 +0800 [thread overview]
Message-ID: <20190829083132.22394-9-duyuyang@gmail.com> (raw)
In-Reply-To: <20190829083132.22394-1-duyuyang@gmail.com>
Given a dependency <prev> -> <next>, two checks are performed:
1. Lock inversion deadlock:
We search whether there is a path from <next> to <prev> in the dependency
graph and if so we have a potential deadlock scenario in check_deadlock_graph().
But if the direct dependency <prev> -> <next> is already in the graph, there
can't be such a path (i.e., <next> to <prev>) because otherwise this path
would have been found when adding the last critical dependency that
completes the circle.
2. IRQ usage violation:
The IRQ usage check searches whether there is a path through <prev> to
<next> that connects an irq-safe lock to an irq-unsafe lock in the
dependency graph in check_irq_usage(). Similarly, if <prev> -> <next> is
already in the graph, there can't be such a path either.
This check skipping should be able to greatly improve performance by
reducing the number of deadlock and IRQ usage checks. This number precisely
equals nr_redundant, which actually is not a small number.
Signed-off-by: Yuyang Du <duyuyang@gmail.com>
---
kernel/locking/lockdep.c | 34 +++++++++++++++++++---------------
1 file changed, 19 insertions(+), 15 deletions(-)
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 4838c99..de088da 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -2433,6 +2433,25 @@ static inline void inc_chains(void)
}
/*
+ * Is the <prev> -> <next> dependency already present?
+ *
+ * (this may occur even though this is a new chain: consider
+ * e.g. the L1 -> L2 -> L3 -> L4 and the L5 -> L1 -> L2 -> L3
+ * chains - the second one will be new, but L1 already has
+ * L2 added to its dependency list, due to the first chain.)
+ */
+ list_for_each_entry(entry, &hlock_class(prev)->locks_after, entry) {
+ if (entry->class == hlock_class(next)) {
+ debug_atomic_inc(nr_redundant);
+
+ if (distance == 1)
+ entry->distance = 1;
+
+ return 1;
+ }
+ }
+
+ /*
* Prove that the new <prev> -> <next> dependency would not
* create a deadlock scenario in the graph. (We do this by
* a breadth-first search into the graph starting at <next>,
@@ -2459,21 +2478,6 @@ static inline void inc_chains(void)
*/
if (next->read == 2 || prev->read == 2)
return 1;
- /*
- * Is the <prev> -> <next> dependency already present?
- *
- * (this may occur even though this is a new chain: consider
- * e.g. the L1 -> L2 -> L3 -> L4 and the L5 -> L1 -> L2 -> L3
- * chains - the second one will be new, but L1 already has
- * L2 added to its dependency list, due to the first chain.)
- */
- list_for_each_entry(entry, &hlock_class(prev)->locks_after, entry) {
- if (entry->class == hlock_class(next)) {
- if (distance == 1)
- entry->distance = 1;
- return 1;
- }
- }
if (!*trace) {
*trace = save_trace();
--
1.8.3.1
next prev parent reply other threads:[~2019-08-29 8:32 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-29 8:31 [PATCH v4 00/30] Support recursive-read lock deadlock detection Yuyang Du
2019-08-29 8:31 ` [PATCH v4 01/30] locking/lockdep: Rename deadlock check functions Yuyang Du
2019-08-29 8:31 ` [PATCH v4 02/30] locking/lockdep: Change return type of add_chain_cache() Yuyang Du
2019-08-29 8:31 ` [PATCH v4 03/30] locking/lockdep: Change return type of lookup_chain_cache_add() Yuyang Du
2019-08-29 8:31 ` [PATCH v4 04/30] locking/lockdep: Pass lock chain from validate_chain() to check_prev_add() Yuyang Du
2019-08-29 8:31 ` [PATCH v4 05/30] locking/lockdep: Add lock chain list_head field in struct lock_list and lock_chain Yuyang Du
2019-08-29 8:31 ` [PATCH v4 06/30] locking/lockdep: Update comments in struct lock_list and held_lock Yuyang Du
2019-08-29 8:31 ` [PATCH v4 07/30] locking/lockdep: Remove indirect dependency redundancy check Yuyang Du
2019-08-29 8:31 ` Yuyang Du [this message]
2019-08-29 8:31 ` [PATCH v4 09/30] locking/lockdep: Remove chain_head argument in validate_chain() Yuyang Du
2019-08-29 8:31 ` [PATCH v4 10/30] locking/lockdep: Remove useless lock type assignment Yuyang Du
2019-08-29 8:31 ` [PATCH v4 11/30] locking/lockdep: Remove irq-safe to irq-unsafe read check Yuyang Du
2019-08-29 8:31 ` [PATCH v4 12/30] locking/lockdep: Specify the depth of current lock stack in lookup_chain_cache_add() Yuyang Du
2019-08-29 8:31 ` [PATCH v4 13/30] locking/lockdep: Treat every lock dependency as in a new lock chain Yuyang Du
2019-08-29 8:31 ` [PATCH v4 14/30] locking/lockdep: Combine lock_lists in struct lock_class into an array Yuyang Du
2019-08-29 8:31 ` [PATCH v4 15/30] locking/lockdep: Consolidate forward and backward lock_lists into one Yuyang Du
2019-08-29 8:31 ` [PATCH v4 16/30] locking/lockdep: Add lock chains to direct lock dependency graph Yuyang Du
2019-08-29 8:31 ` [PATCH v4 17/30] locking/lockdep: Use lock type enum to explicitly specify read or write locks Yuyang Du
2019-08-29 8:31 ` [PATCH v4 18/30] ocking/lockdep: Add read-write type for a lock dependency Yuyang Du
2019-08-29 8:31 ` [PATCH v4 19/30] locking/lockdep: Add helper functions to operate on the searched path Yuyang Du
2019-08-29 8:31 ` [PATCH v4 20/30] locking/lockdep: Update direct dependency's read-write type if it exists Yuyang Du
2019-08-29 8:31 ` [PATCH v4 21/30] locking/lockdep: Introduce chain_hlocks_type for held lock's read-write type Yuyang Du
2019-08-29 8:31 ` [PATCH v4 22/30] locking/lockdep: Hash held lock's read-write type into chain key Yuyang Du
2019-08-29 8:31 ` [PATCH v4 23/30] locking/lockdep: Adjust BFS algorithm to support multiple matches Yuyang Du
2019-08-29 8:31 ` [PATCH v4 24/30] locking/lockdep: Define the two task model for lockdep checks formally Yuyang Du
2019-08-29 8:31 ` [PATCH v4 25/30] locking/lockdep: Introduce mark_lock_unaccessed() Yuyang Du
2019-08-29 8:31 ` [PATCH v4 26/30] locking/lockdep: Add nest lock type Yuyang Du
2019-08-29 8:31 ` [PATCH v4 27/30] locking/lockdep: Add lock exclusiveness table Yuyang Du
2019-08-29 8:31 ` [PATCH v4 28/30] locking/lockdep: Support read-write lock's deadlock detection Yuyang Du
2019-08-29 8:31 ` [PATCH v4 29/30] locking/lockdep: Adjust selftest case for recursive read lock Yuyang Du
2019-08-29 8:31 ` [PATCH v4 30/30] locking/lockdep: Add more lockdep selftest cases Yuyang Du
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190829083132.22394-9-duyuyang@gmail.com \
--to=duyuyang@gmail.com \
--cc=boqun.feng@gmail.com \
--cc=bvanassche@acm.org \
--cc=frederic@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=longman@redhat.com \
--cc=ming.lei@redhat.com \
--cc=mingo@kernel.org \
--cc=paulmck@linux.vnet.ibm.com \
--cc=peterz@infradead.org \
--cc=tglx@linutronix.de \
--cc=will.deacon@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).