mm/damon: add optional debugging-purpose sanity checks
Subsystems (3)
SeongJae Park
maintainerv2 cover mm/damon: add optional debugging-purpose sanity checks
DAMON code has a few assumptions that can be critical if violated. Validating the assumptions in code can be useful at finding such critical bugs. I was actually adding some such additional sanity checks in my personal tree, and those were useful at finding bugs that I made during the development of new patches. We also found [1] sometimes the assumptions are misunderstood. The validation can work as good documentation for such cases.
Add some of such debugging purpose sanity checks. Because those additional checks can impose more overhead, make those only optional via new config, CONFIG_DAMON_DEBUG_SANITY, that is recommended for only development and test setups. And as recommended, enable it for DAMON kunit tests and selftests.
Note that the verification only WARN_ON() for each of the insanity. The developer or tester may better to set panic_on_oops together, like damon-tests/corr did [2].
[1] https://lore.kernel.org/[email protected]
[2] https://github.com/damonitor/damon-tests/commit/a80fbee55e272f151b4e5809ee85898aea33e6ff
Changes from RFC v1
(https://lore.kernel.org/[email protected])
- Fix WARN_ON() failure on kunit UML build
- Fix Wformat-zero-length issue
(https://lore.kernel.org/oe-kbuild-all/[email protected]/)
- Rebase to latest mm-new
- Make warning message less verbose
- Use WARN_ONCE() condition
- Do the verification asap
- Minor cleanups
SeongJae Park (10):
mm/damon: add CONFIG_DAMON_DEBUG_SANITY mm/damon/core: add damon_new_region() debug_sanity check mm/damon/core: add damon_del_region() debug_sanity check mm/damon/core: add damon_nr_regions() debug_sanity check mm/damon/core: add damon_merge_two_regions() debug_sanity check mm/damon/core: add damon_merge_regions_of() debug_sanity check mm/damon/core: add damon_split_region_at() debug_sanity check mm/damon/core: add damon_reset_aggregated() debug_sanity check mm/damon/tests/.kunitconifg: enable DAMON_DEBUG_SANITY selftests/damon/config: enable DAMON_DEBUG_SANITY
mm/damon/Kconfig | 11 +++ mm/damon/core.c | 109 +++++++++++++++++++++++++++ mm/damon/tests/.kunitconfig | 3 + tools/testing/selftests/damon/config | 1 + 4 files changed, 124 insertions(+)
base-commit: b511f483f5fc7e85ad60679181503f16937feb6e
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 1/10 mm/damon: add CONFIG_DAMON_DEBUG_SANITY
Add a new build config that will enable additional DAMON sanity checks. It is recommended to be enabled on only development and test setups, since it can impose additional overhead.
--- mm/damon/Kconfig | 11 +++++++++++ 1 file changed, 11 insertions(+)
mm/damon/Kconfig +11 -0
@@ -12,6 +12,17 @@ config DAMON See https://www.kernel.org/doc/html/latest/mm/damon/index.html for more information. +config DAMON_DEBUG_SANITY + bool "Check sanity of DAMON code" + depends on DAMON + help + This enables additional DAMON debugging-purpose sanity checks in + DAMON code. This can be useful for finding bugs, but impose + additional overhead. This is therefore recommended to be enabled on + only development and test setups. + + If unsure, say N. + config DAMON_KUNIT_TEST bool "Test for damon" if !KUNIT_ALL_TESTS depends on DAMON && KUNIT=y
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 2/10 mm/damon/core: add damon_new_region() debug_sanity check
In reply to SeongJae Parkdamon_new_region() is supposed to be called with only valid address range arguments. Do the check under DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
mm/damon/core.c +12 -0
@@ -109,6 +109,17 @@ int damon_select_ops(struct damon_ctx *ctx, enum damon_ops_id id)return err; } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_new_region (unsigned long start,unsigned long end) +{ +WARN_ONCE (start >= end,"start %lu >= end %lu\n" , start, end); +} +#else +static void damon_verify_new_region (unsigned long start,unsigned long end) +{ +} +#endif +/* * Construct a damon_region struct *
@@ -118,6 +129,7 @@ struct damon_region *damon_new_region(unsigned long start, unsigned long end) {struct damon_region *region; +damon_verify_new_region (start, end); region =kmem_cache_alloc (damon_region_cache,GFP_KERNEL );if (!region)return NULL ;
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 3/10 mm/damon/core: add damon_del_region() debug_sanity check
In reply to SeongJae Parkdamon_del_region() should be called for targets that have one or more regions. Add a sanity check for that under CONFIG_DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 13 +++++++++++++ 1 file changed, 13 insertions(+)
mm/damon/core.c +13 -0
@@ -152,8 +152,21 @@ void damon_add_region(struct damon_region *r, struct damon_target *t) t->nr_regions++; } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_del_region (struct damon_target *t) +{ +WARN_ONCE (t->nr_regions ==0 ,"t->nr_regions == 0\n" ); +} +#else +static void damon_verify_del_region (struct damon_target *t) +{ +} +#endif +static void damon_del_region (struct damon_region *r,struct damon_target *t) { +damon_verify_del_region (t); +list_del (&r->list); t->nr_regions--; }
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 4/10 mm/damon/core: add damon_nr_regions() debug_sanity check
In reply to SeongJae Parkdamon_target->nr_regions is introduced to get the number quickly without having to iterate regions always. Add a sanity check for that under CONFIG_DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
mm/damon/core.c +19 -0
@@ -545,8 +545,27 @@ void damon_destroy_target(struct damon_target *t, struct damon_ctx *ctx)damon_free_target (t); } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_nr_regions (struct damon_target *t) +{ +struct damon_region *r; +unsigned int count =0 ; + +damon_for_each_region (r, t) + count++; +WARN_ONCE (count != t->nr_regions,"t->nr_regions (%u) != count (%u)\n" , + t->nr_regions, count); +} +#else +static void damon_verify_nr_regions (struct damon_target *t) +{ +} +#endif +unsigned int damon_nr_regions (struct damon_target *t) { +damon_verify_nr_regions (t); +return t->nr_regions; }
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 5/10 mm/damon/core: add damon_merge_two_regions() debug_sanity check
In reply to SeongJae ParkA data corruption could cause damon_merge_two_regions() creating zero length DAMON regions. Add a sanity check for that under CONFIG_DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+)
mm/damon/core.c +16 -0
@@ -2473,6 +2473,21 @@ static void kdamond_apply_schemes(struct damon_ctx *c)mutex_unlock (&c->walk_control_lock); } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_merge_two_regions ( +struct damon_region *l,struct damon_region *r) +{ +/* damon_merge_two_regions() may created incorrect left region */ +WARN_ONCE (l->ar.start >= l->ar.end,"l: %lu-%lu, r: %lu-%lu\n" , + l->ar.start, l->ar.end, r->ar.start, r->ar.end); +} +#else +static void damon_verify_merge_two_regions ( +struct damon_region *l,struct damon_region *r) +{ +} +#endif +/* * Merge two adjacent regions into one region */
@@ -2486,6 +2501,7 @@ static void damon_merge_two_regions(struct damon_target *t, l->nr_accesses_bp = l->nr_accesses *10000 ; l->age = (l->age * sz_l + r->age * sz_r) / (sz_l + sz_r); l->ar.end = r->ar.end; +damon_verify_merge_two_regions (l, r);damon_destroy_region (r, t); }
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 6/10 mm/damon/core: add damon_merge_regions_of() debug_sanity check
In reply to SeongJae Parkdamon_merge_regions_of() should be called only after aggregation is finished and therefore each region's nr_accesses and nr_accesses_bp match. There were bugs that broke the assumption, during development of online DAMON parameter updates and monitoring results handling changes. Add a sanity check for that under CONFIG_DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+)
mm/damon/core.c +15 -0
@@ -2505,6 +2505,20 @@ static void damon_merge_two_regions(struct damon_target *t,damon_destroy_region (r, t); } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_merge_regions_of (struct damon_region *r) +{ +WARN_ONCE (r->nr_accesses != r->nr_accesses_bp /10000 , +"nr_accesses (%u) != nr_accesses_bp (%u)\n" , + r->nr_accesses, r->nr_accesses_bp); +} +#else +static void damon_verify_merge_regions_of (struct damon_region *r) +{ +} +#endif + +/* * Merge adjacent regions having similar access frequencies *
@@ -2518,6 +2532,7 @@ static void damon_merge_regions_of(struct damon_target *t, unsigned int thres,struct damon_region *r, *prev =NULL , *next;damon_for_each_region_safe (r, next, t) { +damon_verify_merge_regions_of (r);if (abs (r->nr_accesses - r->last_nr_accesses) > thres) r->age =0 ;else if ((r->nr_accesses ==0 ) != (r->last_nr_accesses ==0 ))
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 7/10 mm/damon/core: add damon_split_region_at() debug_sanity check
In reply to SeongJae Parkdamon_split_region_at() should be called with the correct address to split on. Add a sanity check for that under CONFIG_DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+)
mm/damon/core.c +16 -0
@@ -2586,6 +2586,21 @@ static void kdamond_merge_regions(struct damon_ctx *c, unsigned int threshold, threshold / 2 < max_thres); } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_split_region_at (struct damon_region *r, +unsigned long sz_r) +{ +WARN_ONCE (sz_r ==0 || sz_r >=damon_sz_region (r), +"sz_r: %lu r: %lu-%lu (%lu)\n" , + sz_r, r->ar.start, r->ar.end,damon_sz_region (r)); +} +#else +static void damon_verify_split_region_at (struct damon_region *r, +unsigned long sz_r) +{ +} +#endif +/* * Split a region in two *
@@ -2597,6 +2612,7 @@ static void damon_split_region_at(struct damon_target *t, {struct damon_region *new; +damon_verify_split_region_at (r, sz_r); new =damon_new_region (r->ar.start + sz_r, r->ar.end);if (!new)return ;
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 8/10 mm/damon/core: add damon_reset_aggregated() debug_sanity check
In reply to SeongJae ParkAt time of damon_reset_aggregated(), aggregation of the interval should be completed, and hence nr_accesses and nr_accesses_bp should match. I found a few bugs caused it to be broken in the past, from online parameters update and complicated nr_accesses handling changes. Add a sanity check for that under CONFIG_DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
mm/damon/core.c +18 -0
@@ -1666,6 +1666,23 @@ static void damon_warn_fix_nr_accesses_corruption(struct damon_region *r) r->nr_accesses_bp = r->nr_accesses *10000 ; } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_reset_aggregated (struct damon_region *r, +struct damon_ctx *c) +{ +WARN_ONCE (r->nr_accesses_bp != r->last_nr_accesses *10000 , +"nr_accesses_bp %u last_nr_accesses %u sis %lu %lu\n" , + r->nr_accesses_bp, r->last_nr_accesses, + c->passed_sample_intervals, c->next_aggregation_sis); +} +#else +static void damon_verify_reset_aggregated (struct damon_region *r, +struct damon_ctx *c) +{ +} +#endif + +/* * Reset the aggregated monitoring results ('nr_accesses' of each region). */
@@ -1682,6 +1699,7 @@ static void kdamond_reset_aggregated(struct damon_ctx *c)damon_warn_fix_nr_accesses_corruption (r); r->last_nr_accesses = r->nr_accesses; r->nr_accesses =0 ; +damon_verify_reset_aggregated (r, c); } ti++; }
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 9/10 mm/damon/tests/.kunitconifg: enable DAMON_DEBUG_SANITY
In reply to SeongJae ParkCONFIG_DAMON_DEBUG_SANITY is recommended for DAMON development and test setups. Enable it on the default configurations for DAMON kunit test run.
--- mm/damon/tests/.kunitconfig | 3 +++ 1 file changed, 3 insertions(+)
mm/damon/tests/.kunitconfig +3 -0
@@ -13,3 +13,6 @@ CONFIG_DAMON_VADDR_KUNIT_TEST=y CONFIG_SYSFS=y CONFIG_DAMON_SYSFS=y CONFIG_DAMON_SYSFS_KUNIT_TEST=y + +# enable DAMON_DEBUG_SANITY to catch any bug +CONFIG_DAMON_DEBUG_SANITY=y
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH v2 10/10 selftests/damon/config: enable DAMON_DEBUG_SANITY
In reply to SeongJae ParkCONFIG_DAMON_DEBUG_SANITY is recommended for DAMON development and test setups. Enable it on the build config for DAMON selftests.
--- tools/testing/selftests/damon/config | 1 + 1 file changed, 1 insertion(+)
tools/testing/selftests/damon/config +1 -0
@@ -4,3 +4,4 @@ CONFIG_DAMON_PADDR=y CONFIG_DAMON_VADDR=y CONFIG_DAMON_RECLAIM=y CONFIG_DAMON_LRU_SORT=y +CONFIG_DAMON_DEBUG_SANITY=y
Signature · 1 line
2.47.3