locking/lockdep: make chain-hlocks average depth configurable
Subsystems (2)
PATCH 1/1 locking/lockdep: make chain-hlocks average depth configurable
The static chain_hlocks[] pool is sized as:
MAX_LOCKDEP_CHAIN_HLOCKS = MAX_LOCKDEP_CHAINS * AVG_LOCKDEP_CHAIN_DEPTH
MAX_LOCKDEP_CHAINS is already tunable via CONFIG_LOCKDEP_CHAINS_BITS, but AVG_LOCKDEP_CHAIN_DEPTH is hardcoded to 5 and has no Kconfig knob. Workloads that build unusually deep individual lock chains -- rather than simply a large number of distinct chains -- can exhaust chain_hlocks[] and trip:
BUG: MAX_LOCKDEP_CHAIN_HLOCKS too low!
well before MAX_LOCKDEP_CHAINS itself is anywhere near full, which silently disables lock debugging for the rest of the boot (debug_locks_off_graph_unlock()). Bumping LOCKDEP_CHAINS_BITS alone does not help in that case since the bottleneck is chain depth, not chain count.
Observed on a PREEMPT_DYNAMIC + RCU lockdep + KASAN debug build, reproducing at every boot from add_chain_cache() failing to allocate out of the static pool, hit from sched wakeup, hrtimer, and btrfs flush-workqueue paths (deep IRQ/softirq nesting stacked on top of deep filesystem/scheduler call chains).
Add CONFIG_LOCKDEP_CHAIN_DEPTH so this can be tuned like LOCKDEP_BITS/LOCKDEP_CHAINS_BITS, defaulting to 5 to preserve current behavior for everyone who isn't hitting this.
--- kernel/locking/lockdep_internals.h | 2 +- lib/Kconfig.debug | 17 +++++++++++++++++ 2 files changed, 18 insertions(+), 1 deletion(-)
kernel/locking/lockdep_internals.h +1 -1
@@ -121,7 +121,7 @@ enum {#define MAX_LOCKDEP_CHAINS (1UL << MAX_LOCKDEP_CHAINS_BITS) -#define AVG_LOCKDEP_CHAIN_DEPTH 5 +#define AVG_LOCKDEP_CHAIN_DEPTH CONFIG_LOCKDEP_CHAIN_DEPTH#define MAX_LOCKDEP_CHAIN_HLOCKS (MAX_LOCKDEP_CHAINS * AVG_LOCKDEP_CHAIN_DEPTH)extern struct lock_chain lock_chains[];
lib/Kconfig.debug +17 -0
@@ -1614,6 +1614,23 @@ config LOCKDEP_CHAINS_BITS help Try increasing this value if you hit "BUG: MAX_LOCKDEP_CHAINS too low!" message. +config LOCKDEP_CHAIN_DEPTH + int "Average depth for MAX_LOCKDEP_CHAIN_HLOCKS" + depends on LOCKDEP + range 3 20 + default 5 + help + Average per-chain depth used to size the static chain_hlocks[] + pool: MAX_LOCKDEP_CHAIN_HLOCKS = MAX_LOCKDEP_CHAINS * + LOCKDEP_CHAIN_DEPTH. + + Workloads that build unusually deep lock chains (heavy irq/softirq + nesting stacked on top of deep filesystem or scheduler call chains) + can exhaust this pool and trip "BUG: MAX_LOCKDEP_CHAIN_HLOCKS too + low!" well before MAX_LOCKDEP_CHAINS itself is exhausted, silently + disabling lock debugging. Increase this value if you hit that + message and LOCKDEP_CHAINS_BITS increases alone don't help. + config LOCKDEP_STACK_TRACE_BITS int "Size for MAX_STACK_TRACE_ENTRIES (as Nth power of 2)" depends on LOCKDEP && !LOCKDEP_SMALL
Signature · 1 line
2.55.0
From: Zhan Xusheng <[email protected]>
Quoted 1 line · Muhammad Bilal
lib/Kconfig.debug +1
-0+ range 3 20
[...]
Review
Quoted 2 lines
(quoted diff) +2
-0+ Increase this value if you hit that message and LOCKDEP_CHAINS_BITS + increases alone don't help.
I might be missing something, but doesn't the top of this range interact with LOCKDEP_CHAINS_BITS in a way that can fail the build?
MAX_LOCKDEP_CHAIN_HLOCKS is (1 << LOCKDEP_CHAINS_BITS) * LOCKDEP_CHAIN_DEPTH, and add_chain_cache() has:
BUILD_BUG_ON((1UL << 24) <= ARRAY_SIZE(chain_hlocks));
since lock_chain.base is a 24-bit index into chain_hlocks[]. So the product has to stay below 2^24.
LOCKDEP_CHAINS_BITS ranges up to 21, so with the new knob:
CHAINS_BITS=21, DEPTH=8 -> 2^21 * 8 == 2^24 -> BUILD_BUG_ON fires
i.e. at the maximum CHAINS_BITS anything above DEPTH=7 stops building. Today, with the fixed depth of 5, 2^21 * 5 stays under the limit, so this is a new combination the patch makes reachable.
What I find a little awkward is that this is exactly the path the help text points people at: someone who already raised LOCKDEP_CHAINS_BITS to 21 chasing the "too low" message and then bumps LOCKDEP_CHAIN_DEPTH as suggested gets a (fairly opaque) BUILD_BUG_ON rather than a larger pool.
Kconfig can't easily express the (1 << CHAINS_BITS) * DEPTH < 2^24 product, so would it make sense to at least spell the constraint out in the help text and note that the build fails otherwise? Capping the range for the worst case (CHAINS_BITS=21 => DEPTH <= 7) would be safe but would needlessly restrict the common CHAINS_BITS=16 case, where much larger depths are fine.
Thanks,
Zhan Xusheng
Good catch! Just send v2 with fix.
Thanks,
Muhammad Bilal
Quoted 41 lines · Zhan Xusheng
From: Zhan Xusheng <[email protected]>
Quoted 1 line · Muhammad Bilal
+ range 3 20
[...]
Review
Quoted 2 lines
(quoted diff) +2
-0+ Increase this value if you hit that message and LOCKDEP_CHAINS_BITS + increases alone don't help.I might be missing something, but doesn't the top of this range interact with
LOCKDEP_CHAINS_BITSin a way that can fail the build?
MAX_LOCKDEP_CHAIN_HLOCKSis (1 <<LOCKDEP_CHAINS_BITS) *LOCKDEP_CHAIN_DEPTH, andadd_chain_cache()has:BUILD_BUG_ON((1UL << 24) <= ARRAY_SIZE(chain_hlocks));since lock_chain.base is a 24-bit index into chain_hlocks[]. So the product has to stay below 2^24.
LOCKDEP_CHAINS_BITSranges up to 21, so with the new knob:CHAINS_BITS=21, DEPTH=8 -> 2^21 * 8 == 2^24 -> BUILD_BUG_ON firesi.e. at the maximum
CHAINS_BITSanything above DEPTH=7 stops building. Today, with the fixed depth of 5, 2^21 * 5 stays under the limit, so this is a new combination the patch makes reachable.What I find a little awkward is that this is exactly the path the help text points people at: someone who already raised
LOCKDEP_CHAINS_BITSto 21 chasing the "too low" message and then bumpsLOCKDEP_CHAIN_DEPTHas suggested gets a (fairly opaque)BUILD_BUG_ONrather than a larger pool.
Kconfigcan't easily express the (1 <<CHAINS_BITS) * DEPTH < 2^24 product, so would it make sense to at least spell the constraint out in the help text and note that the build fails otherwise? Capping the range for the worst case (CHAINS_BITS=21 => DEPTH <= 7) would be safe but would needlessly restrict the common CHAINS_BITS=16 case, where much larger depths are fine.Thanks,
Zhan Xusheng