mm/damon: minor improvements for code readability and tests
Subsystems (3)
SeongJae Park
maintainercover mm/damon: minor improvements for code readability and tests
Implement minor improvements on code readability and tests for DAMON.
First seven patches are for DAMON code readability and resulting maintenance. Patches 1 and 2 make damon_set_regions() safer and easier to read. Patches 3 and 4 remove fragmented DAMON API use cases. Patches 5-7 hides unused core functions that are unnecessarily exposed to API callers.
The following seven patches are for DAMON tests improvement. Patches 8 and 9 adds and removes DAMON_DEBUG_SANITY verifications to ensure reasonable test coverage without too high overhead. Patch 10 adds a new kunit test for damon_set_regions(). Patch 11 makes sysfs.py selftest more gracefully finishes under test failures. Patches 12-13 adds simple sysfs.sh test cases for the monitoring intervals goal directory, the addr_unit file and the pause file.
Changes from RFC v1.1
- RFC v1.1: https://lore.kernel.org/[email protected]
- Free DAMON target in vaddr unit test fail-out path. Changes from RFC v1
- RFC v1: https://lore.kernel.org/[email protected]
- Handle damon_set_regions() failure in vaddr unit test.
- Free ranges array in vaddr unit test.
- Fix wrong region address verification.
- Fix typos in selftest: s/exit/exist/
SeongJae Park (14):
mm/damon/core: safely handle no region case in damon_set_regions() mm/damon/core: do not use region out of a loop in damon_set_regions() samples/damon/mtier: replace damon_add_region() with
damon_set_regions()
mm/damon/tests/vaddr-kunit: replace damon_add_region() with
damon_set_regions()
mm/damon/core: hide damon_add_region()
mm/damon/core: hide damon_insert_region()
mm/damon/core: hide damon_destroy_region()
mm/damon/core: add kdamond_call() debug_sanity check mm/damon/core: remove damon_verify_nr_regions() mm/damon/tests/core-kunit: add damon_set_regions() test cases selftests/damon/sysfs.py: stop kdamonds before failing selftests/damon/sysfs.sh: test monitoring intervals goal dir selftests/damon/sysfs.sh: test addr_unit file existence selftests/damon/sysfs.sh: test pause file existence
include/linux/damon.h | 13 --- mm/damon/core.c | 92 ++++++++++++---- mm/damon/tests/core-kunit.h | 142 +++++++++++++++++++++---- mm/damon/tests/vaddr-kunit.h | 27 +++-- samples/damon/mtier.c | 10 +- tools/testing/selftests/damon/sysfs.py | 4 + tools/testing/selftests/damon/sysfs.sh | 14 +++ 7 files changed, 233 insertions(+), 69 deletions(-)
base-commit: ac6b83c7bf3671e8ff0d541cf03e6eaeed00a79e
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 1/14 mm/damon/core: safely handle no region case in damon_set_regions()
damon_set_regions() calls damon_first_region() regardless of the number of DAMON regions in a given DAMON target. damon_first_region() internally uses list_first_entry(), which clearly documents the list is expected to be not empty. Due to the internal implementation of the macro, damon_set_regions() is safe for now. But the internal implementation of the macro can be changed in future. Refactor the function to explicitly and safely handle the empty region list case without depending on the internal implementation.
No behavioral change is intended.
--- mm/damon/core.c | 13 +++++++++++++ 1 file changed, 13 insertions(+)
mm/damon/core.c +13 -0
@@ -356,6 +356,19 @@ int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges,damon_destroy_region (r, t); } +if (!damon_nr_regions (t)) { +for (i =0 ; i < nr_ranges; i++) { + r =damon_new_region ( +ALIGN_DOWN (ranges[i].start, + min_region_sz), +ALIGN (ranges[i].end, min_region_sz)); +if (!r) +return -ENOMEM ; +damon_add_region (r, t); + } +return 0 ; + } + r =damon_first_region (t);/* Add new regions or resize existing regions to fit in the ranges */ for (i =0 ; i < nr_ranges; i++) {
Signature · 1 line
2.47.3
[email protected]
AI reviewRe: [RFC PATCH v1.2 01/14] mm/damon/core: safely handle no region case in damon_set_regions()
Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:
Pre-existing issues:
- [Critical] The ALIGN() macro can overflow when processing malicious or overly large user-provided region 'end' values, creating invalid regions where start > end.
- [High] min_region_sz can be a non-power-of-2 value, which corrupts region boundary calculations when passed to ALIGN() and ALIGN_DOWN().
- [High] Userspace can maliciously trigger a kernel warning (WARN_ONCE) by providing a region where start == end via sysfs.
- [High] A memory allocation failure during region creation results in a partial state update without a rollback, potentially leaving the running DAMON context corrupted. --
commit fa588352eaf7b151171cca8c2acb0bb710a71c75
Author: SeongJae Park <[email protected]>
mm/damon/core: safely handle no region case in damon_set_regions()
This commit refactors damon_set_regions() to explicitly handle the empty region list case without depending on the internal implementation of list_first_entry().
mm/damon/core.c int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, · line 363
Quoted 13 lines
mm/damon/core.c +5
-0@@ -356,6 +356,19 @@ int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges,damon_destroy_region (r, t); } +if (!damon_nr_regions (t)) { +for (i =0 ; i < nr_ranges; i++) { + r = damon_new_region( + ALIGN_DOWN(ranges[i].start, + min_region_sz),
[Severity: High]
This is a pre-existing issue, but can min_region_sz be a non-power-of-2 value here?
If sys_ctx->addr_unit is set to an arbitrary value via sysfs, min_region_sz might not be a power of 2. The ALIGN() and ALIGN_DOWN() macros internally perform bitwise AND operations with ~(a - 1), which assumes the alignment is a power of 2.
Could passing a non-power-of-2 value truncate arbitrary bits and result in misaligned boundaries where start > end?
mm/damon/core.c int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, · line 364
Quoted 1 line
mm/damon/core.c +1
-0+ ALIGN(ranges[i].end, min_region_sz));
[Severity: Critical]
This is a pre-existing issue, but can ALIGN() overflow when processing overly large end values?
If a user passes ULONG_MAX as the region end via sysfs, ALIGN(ULONG_MAX, 4096) will overflow and wrap around to 0.
Would this result in damon_new_region(start, 0) being called, creating an invalid region where start > end, which could underflow the size and cause out-of-bounds memory accesses?
[Severity: High]
This isn't a bug introduced by this patch, but what happens if userspace provides a region where start == end?
The sysfs interface end_store accepts arbitrary values, and if start == end, both bypass the start > end validation check. Since ALIGN_DOWN() and ALIGN() return identical values for identical inputs, would damon_new_region() be invoked with equal start and end arguments, allowing an attacker with sysfs access to maliciously trigger a WARN_ONCE(start >= end)?
mm/damon/core.c int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, · line 366
Quoted 2 lines
mm/damon/core.c +2
-0+if (!r) +return -ENOMEM ;
[Severity: High]
This is a pre-existing issue, but does this allocation failure leave the DAMON context in a corrupted state?
If damon_new_region() returns NULL due to memory pressure, the function immediately returns -ENOMEM without rolling back the regions that were already removed or added in the previous loop.
Could this leave the running kdamond thread with a partially modified list of regions?
Quoted 6 lines
mm/damon/core.c +5
-0+damon_add_region (r, t); + } +return 0 ; + } + r =damon_first_region (t);
[ ... ]
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
SeongJae Park
maintainerRe: [RFC PATCH v1.2 01/14] mm/damon/core: safely handle no region case in damon_set_regions()
Quoted 7 lines · On Thu, 21 May 2026 15:19:28 +0000 [email protected] wrote:
Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:
Pre-existing issues:
- [Critical] TheALIGN()macro can overflow when processing malicious or overly large user-provided region 'end' values, creating invalid regions where start > end.
- [High]min_region_szcan be a non-power-of-2 value, which corrupts region boundary calculations when passed toALIGN()andALIGN_DOWN().
- [High] Userspace can maliciously trigger a kernel warning (WARN_ONCE) by providing a region where start == end via sysfs.
- [High] A memory allocation failure during region creation results in a partial state update without a rollback, potentially leaving the running DAMON context corrupted.
No one is caused by this patch, so no blocker for this patch.
The last finding seems worthy to separately work for. I will work on it.
Please read below for more details.
Quoted 29 lines
--
commit
fa588352eaf7b151171cca8c2acb0bb710a71c75
Author: SeongJae Park <[email protected]>
mm/damon/core: safely handle no region case indamon_set_regions()This commit refactors
damon_set_regions()to explicitly handle the empty region list case without depending on the internal implementation oflist_first_entry().mm/damon/core.c int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, · line 363
Quoted 13 lines
mm/damon/core.c +5
-0@@ -356,6 +356,19 @@ int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges,damon_destroy_region (r, t); } +if (!damon_nr_regions (t)) { +for (i =0 ; i < nr_ranges; i++) { + r = damon_new_region( + ALIGN_DOWN(ranges[i].start, + min_region_sz),[Severity: High]
This is a pre-existing issue, but canmin_region_szbe a non-power-of-2 value here?
Quoted 7 lines
If
sys_ctx->addr_unitis set to an arbitrary value via sysfs,min_region_szmight not be a power of 2. TheALIGN()andALIGN_DOWN()macros internally perform bitwise AND operations with ~(a - 1), which assumes the alignment is a power of 2.Could passing a non-power-of-2 value truncate arbitrary bits and result in misaligned boundaries where start > end?
DAMON_SYSFS enforces the power of 2 using damon_commit_ctx() based parameter validation. DAMON_RECLAIM and DAMON_LRU_SORT also do the checks in their ..._apply_parameters() functions. So this cannot happen.
Quoted 7 lines
mm/damon/core.c int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, · line 364
Quoted 2 lines
mm/damon/core.c +1
-0+ ALIGN(ranges[i].end, min_region_sz));[Severity: Critical]
This is a pre-existing issue, but canALIGN()overflow when processing overly large end values?
Quoted 6 lines
If a user passes
ULONG_MAXas the region end via sysfs,ALIGN(ULONG_MAX, 4096)will overflow and wrap around to 0.Would this result in
damon_new_region(start, 0)being called, creating an invalid region where start > end, which could underflow the size and cause out-of-bounds memory accesses?
As I previously answered [1], that input is unlikely in real use case.
Quoted 10 lines
[Severity: High]
This isn't a bug introduced by this patch, but what happens if userspace provides a region where start == end?The sysfs interface
end_storeaccepts arbitrary values, and if start == end, both bypass the start > end validation check. SinceALIGN_DOWN()andALIGN()return identical values for identical inputs, woulddamon_new_region()be invoked with equal start and end arguments, allowing an attacker with sysfs access to maliciously trigger aWARN_ONCE(start >= end)?
Again, unlikely input in real use case. To intentionally to this, the user need sysfs write permission. If a malicious user got the permission, I think they can do anything bad?
Quoted 8 lines
mm/damon/core.c int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, · line 366
Quoted 3 lines
mm/damon/core.c +2
-0+if (!r) +return -ENOMEM ;[Severity: High]
This is a pre-existing issue, but does this allocation failure leave the DAMON context in a corrupted state?
Quoted 6 lines
If
damon_new_region()returns NULL due to memory pressure, the function immediately returns-ENOMEMwithout rolling back the regions that were already removed or added in the previous loop.Could this leave the running kdamond thread with a partially modified list of regions?
Yes, and the caller should do damon_destroy_ctx() of the ctx. And DAMON_SYSFS does that, while vaddr doesn't. I will work on this.
[1] https://lore.kernel.org/[email protected]
Thanks,
SJ
[...]
SeongJae Park
maintainerRe: [RFC PATCH v1.2 01/14] mm/damon/core: safely handle no region case in damon_set_regions()
Quoted 32 lines · On Thu, 21 May 2026 19:16:42 -0700 SeongJae Park <[email protected]> wrote:
Quoted 7 lines · On Thu, 21 May 2026 15:19:28 +0000 [email protected] wrote:
Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:
Pre-existing issues:
- [Critical] TheALIGN()macro can overflow when processing malicious or overly large user-provided region 'end' values, creating invalid regions where start > end.
- [High]min_region_szcan be a non-power-of-2 value, which corrupts region boundary calculations when passed toALIGN()andALIGN_DOWN().
- [High] Userspace can maliciously trigger a kernel warning (WARN_ONCE) by providing a region where start == end via sysfs.
- [High] A memory allocation failure during region creation results in a partial state update without a rollback, potentially leaving the running DAMON context corrupted.No one is caused by this patch, so no blocker for this patch.
The last finding seems worthy to separately work for. I will work on it.
Please read below for more details.
Quoted 16 lines
--
commit
fa588352eaf7b151171cca8c2acb0bb710a71c75
Author: SeongJae Park <[email protected]>
mm/damon/core: safely handle no region case indamon_set_regions()This commit refactors
damon_set_regions()to explicitly handle the empty region list case without depending on the internal implementation oflist_first_entry().Quoted 5 lines
mm/damon/core.c +0
-0@@ -356,6 +356,19 @@ int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges,
[...]
Quoted 11 lines
Quoted 11 lines
mm/damon/core.c int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, · line 366
Quoted 2 lines
mm/damon/core.c +2
-0+if (!r) +return -ENOMEM ;[Severity: High]
This is a pre-existing issue, but does this allocation failure leave the DAMON context in a corrupted state?If
damon_new_region()returns NULL due to memory pressure, the function immediately returns-ENOMEMwithout rolling back the regions that were already removed or added in the previous loop.
Quoted 5 lines
Quoted 2 lines
Could this leave the running kdamond thread with a partially modified list of regions?
Yes, and the caller should do
damon_destroy_ctx()of the ctx. AndDAMON_SYSFSdoes that, while vaddr doesn't. I will work on this.
Hmm, but the vaddr's damon_set_regions() calls are only for the automated mapping detection based monitoring target address rangs setup. In the mode, the monitoring target address ranges setup is only the best effort, and continuously retried for every operation set update interval. Hence, ignoring the error is actually intentional and fine. So this is not a real issue and therefore no fix is needed.
Let me know if I'm missing something, though.
Thanks,
SJ
[...]
SeongJae Park
maintainerRFC PATCH 2/14 mm/damon/core: do not use region out of a loop in damon_set_regions()
In reply to SeongJae Parkdamon_set_regions() assumes the DAMON region iterator is referencing the last region after the region iteration loop is completed. The code is indeed implemented in the way, but that is not a documented safe behavior. Hence it is unreliable and difficult to read. Cleanup the code to avoid the case.
No behavioral change is intended.
--- mm/damon/core.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)
mm/damon/core.c +9 -2
@@ -374,6 +374,7 @@ int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges,for (i =0 ; i < nr_ranges; i++) {struct damon_region *first =NULL , *last, *newr;struct damon_addr_range *range; +bool insert_before_r = false; range = &ranges[i];/* Get the first/last regions intersecting with the range */
@@ -383,8 +384,10 @@ int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, first = r; last = r; } -if (r->ar.start >= range->end) +if (r->ar.start >= range->end) { + insert_before_r = true;break ; + } }if (!first) {/* no region intersects with this range */
@@ -394,7 +397,11 @@ int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges, ALIGN(range->end, min_region_sz));if (!newr)return -ENOMEM ; -damon_insert_region (newr,damon_prev_region (r), r, t); +if (insert_before_r) +damon_insert_region (newr,damon_prev_region (r), + r, t); +else +damon_add_region (newr, t); }else {/* resize intersecting regions to fit in this range */ first->ar.start = ALIGN_DOWN(range->start,
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 3/14 samples/damon/mtier: replace damon_add_region() with damon_set_regions()
In reply to SeongJae Parkmtier DAMON sample module and DAMON virtual address operation set (vaddr) unit tests are using damon_add_region() for setup of DAMON monitoring target region boundaries setup. But, damon_set_regions() is designed for exactly the purpose. All other DAMON API callers use the function for the purpose. Replace damon_add_region() usage in mtier sample module with damon_set_regions(), for unifying the use case and reducing the maintenance cost.
--- samples/damon/mtier.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
samples/damon/mtier.c +6 -4
@@ -75,11 +75,11 @@ static struct damon_ctx *damon_sample_mtier_build_ctx(bool promote)struct damon_ctx *ctx;struct damon_attrs attrs;struct damon_target *target; -struct damon_region *region;struct damos *scheme;struct damos_quota_goal *quota_goal;struct damos_filter *filter;struct region_range addr; +struct damon_addr_range range;int ret; ctx =damon_new_ctx ();
@@ -120,10 +120,12 @@ static struct damon_ctx *damon_sample_mtier_build_ctx(bool promote) addr.end = promote ? node1_end_addr : node0_end_addr; } - region =damon_new_region (addr.start, addr.end); -if (!region) + range.start = addr.start; + range.end = addr.end; + + ret =damon_set_regions (target, &range,1 ,DAMON_MIN_REGION_SZ ); +if (ret) goto free_out; -damon_add_region (region, target); scheme = damon_new_scheme(/* access pattern */
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 4/14 mm/damon/tests/vaddr-kunit: replace damon_add_region() with damon_set_regions()
In reply to SeongJae ParkDAMON virtual address operation set (vaddr) unit tests is using damon_add_region() for setup of DAMON monitoring target region boundaries setup. But, damon_set_regions() is designed for exactly the purpose. All other DAMON API callers use the function for the purpose. Replace damon_add_region() usage in the unit tests with damon_set_regions(), for unifying the use case and reducing the maintenance cost.
--- mm/damon/tests/vaddr-kunit.h | 27 ++++++++++++++++++++------- 1 file changed, 20 insertions(+), 7 deletions(-)
mm/damon/tests/vaddr-kunit.h +20 -7
@@ -132,22 +132,35 @@ static void damon_do_test_apply_three_regions(struct kunit *test, unsigned long *expected, int nr_expected) {struct damon_target *t; +struct damon_addr_range *ranges;struct damon_region *r;int i; t =damon_new_target ();if (!t)kunit_skip (test,"target alloc fail" ); + + ranges =kmalloc_array (nr_regions /2 ,sizeof (*ranges),GFP_KERNEL ); +if (!ranges) { +damon_destroy_target (t,NULL ); +kunit_skip (test,"ranges alloc fail" ); + }for (i =0 ; i < nr_regions /2 ; i++) { - r =damon_new_region (regions[i *2 ], regions[i *2 +1 ]); -if (!r) { -damon_destroy_target (t,NULL ); -kunit_skip (test,"region alloc fail" ); - } -damon_add_region (r, t); + ranges[i].start = regions[i *2 ]; + ranges[i].end = regions[i *2 +1 ]; } +if (damon_set_regions (t, ranges, nr_regions /2 , +DAMON_MIN_REGION_SZ )) { +kfree (ranges); +damon_destroy_target (t,NULL ); +kunit_skip (test,"damon_set_regions() fail" ); + } +kfree (ranges); -damon_set_regions (t, three_regions,3 ,DAMON_MIN_REGION_SZ ); +if (damon_set_regions (t, three_regions,3 ,DAMON_MIN_REGION_SZ )) { +damon_destroy_target (t,NULL ); +kunit_skip (test,"second damon_set_regions() fail" ); + }for (i =0 ; i < nr_expected /2 ; i++) { r =__nth_region_of (t, i);
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 5/14 mm/damon/core: hide damon_add_region()
In reply to SeongJae Parkdamon_add_region() is being used by only DAMON core, but exposed to DAMON API callers. Exposing something that is not really being used by others will only increase the maintenance cost. Hide it.
--- include/linux/damon.h | 1 - mm/damon/core.c | 2 +- 2 files changed, 1 insertion(+), 2 deletions(-)
include/linux/damon.h +0 -1
@@ -1024,7 +1024,6 @@ static inline void damon_insert_region(struct damon_region *r, t->nr_regions++; } -void damon_add_region (struct damon_region *r,struct damon_target *t);void damon_destroy_region (struct damon_region *r,struct damon_target *t);int damon_set_regions (struct damon_target *t,struct damon_addr_range *ranges,unsigned int nr_ranges,unsigned long min_region_sz);
mm/damon/core.c +1 -1
@@ -250,7 +250,7 @@ struct damon_region *damon_new_region(unsigned long start, unsigned long end)return region; } -void damon_add_region (struct damon_region *r,struct damon_target *t) +static void damon_add_region (struct damon_region *r,struct damon_target *t) {list_add_tail (&r->list, &t->regions_list); t->nr_regions++;
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 6/14 mm/damon/core: hide damon_insert_region()
In reply to SeongJae Parkdamon_insert_region() is being used by only DAMON core, but exposed to DAMON API callers. Exposing something that is not really being used by others will only increase the maintenance cost. Hide it.
--- include/linux/damon.h | 11 ----------- mm/damon/core.c | 11 +++++++++++ 2 files changed, 11 insertions(+), 11 deletions(-)
include/linux/damon.h +0 -11
@@ -1013,17 +1013,6 @@ void damon_add_probe(struct damon_ctx *ctx, struct damon_probe *probe);struct damon_region *damon_new_region (unsigned long start,unsigned long end); -/* -* Add a region between two other regions -*/ -static inline void damon_insert_region (struct damon_region *r, -struct damon_region *prev,struct damon_region *next, -struct damon_target *t) -{ -__list_add (&r->list, &prev->list, &next->list); - t->nr_regions++; -} -void damon_destroy_region (struct damon_region *r,struct damon_target *t);int damon_set_regions (struct damon_target *t,struct damon_addr_range *ranges,unsigned int nr_ranges,unsigned long min_region_sz);
mm/damon/core.c +11 -0
@@ -256,6 +256,17 @@ static void damon_add_region(struct damon_region *r, struct damon_target *t) t->nr_regions++; } +/* +* Add a region between two other regions +*/ +static inline void damon_insert_region (struct damon_region *r, +struct damon_region *prev,struct damon_region *next, +struct damon_target *t) +{ +__list_add (&r->list, &prev->list, &next->list); + t->nr_regions++; +} +#ifdef CONFIG_DAMON_DEBUG_SANITY static void damon_verify_del_region (struct damon_target *t) {
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 7/14 mm/damon/core: hide damon_destroy_region()
In reply to SeongJae Parkdamon_destroy_region() is being used by only DAMON core, but exposed to DAMON API callers. Exposing something that is not really being used by others will only increase the maintenance cost. Hide it.
--- include/linux/damon.h | 1 - mm/damon/core.c | 3 ++- 2 files changed, 2 insertions(+), 2 deletions(-)
include/linux/damon.h +0 -1
@@ -1013,7 +1013,6 @@ void damon_add_probe(struct damon_ctx *ctx, struct damon_probe *probe);struct damon_region *damon_new_region (unsigned long start,unsigned long end); -void damon_destroy_region (struct damon_region *r,struct damon_target *t);int damon_set_regions (struct damon_target *t,struct damon_addr_range *ranges,unsigned int nr_ranges,unsigned long min_region_sz); void damon_update_region_access_rate(struct damon_region *r, bool accessed,
mm/damon/core.c +2 -1
@@ -291,7 +291,8 @@ static void damon_free_region(struct damon_region *r)kmem_cache_free (damon_region_cache, r); } -void damon_destroy_region (struct damon_region *r,struct damon_target *t) +static void damon_destroy_region (struct damon_region *r, +struct damon_target *t) {damon_del_region (r, t);damon_free_region (r);
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 8/14 mm/damon/core: add kdamond_call() debug_sanity check
In reply to SeongJae Parkkdamond_call() is the place where DAMON API callers are allowed to access the DAMON context's public internal state including the monitoring results. Hence it is important to ensure it is called with the expected DAMON context state. Do the check under DAMON_DEBUG_SANITY.
--- mm/damon/core.c | 33 +++++++++++++++++++++++++++++++++ 1 file changed, 33 insertions(+)
mm/damon/core.c +33 -0
@@ -3331,6 +3331,37 @@ static void kdamond_usleep(unsigned long usecs)usleep_range_idle (usecs, usecs +1 ); } +#ifdef CONFIG_DAMON_DEBUG_SANITY +static void damon_verify_ctx (struct damon_ctx *c) +{ +struct damon_target *t; +struct damon_region *r; + +damon_for_each_target (t, c) { +struct damon_region *prev_r =NULL ; +unsigned int nr_regions =0 ; + +damon_for_each_region (r, t) { +WARN_ONCE (r->ar.start >= r->ar.end, +"region start (%lu) >= end (%lu)\n" , + r->ar.start, r->ar.end); +WARN_ONCE (prev_r && prev_r->ar.end > r->ar.start, +"region overlap (%lu > %lu)\n" , + prev_r->ar.end, r->ar.start); + prev_r = r; + nr_regions++; + } +WARN_ONCE (damon_nr_regions (t) != nr_regions, +"nr_regions mismatch: %u != %u\n" , +damon_nr_regions (t), nr_regions); + } +} +#else +static void damon_verify_ctx (struct damon_ctx *c) +{ +} +#endif +/* * kdamond_call() - handle damon_call_control objects. * @ctx: The &struct damon_ctx of the kdamond.
@@ -3346,6 +3377,8 @@ static void kdamond_call(struct damon_ctx *ctx, bool cancel)struct damon_call_control *control, *next;LIST_HEAD (controls); +damon_verify_ctx (ctx); +mutex_lock (&ctx->call_controls_lock);list_splice_tail_init (&ctx->call_controls, &controls);mutex_unlock (&ctx->call_controls_lock);
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 9/14 mm/damon/core: remove damon_verify_nr_regions()
In reply to SeongJae ParkWhen CONFIG_DAMON_DEBUG_SANITY is enabled, damon_verify_nr_regions() is called for each damon_nr_regions() invocation. damon_veify_nr_regions() iterates all regions. damon_nr_regions() is called for each region in kdamond_reset_aggregated() and damos_apply_scheme(). Hence it imposes O(n**2) overhead where n is the number of regions.
Though the verification is enabled only under DAMON_DEBUG_SANITY, which is not for production use cases, it could be too high overhead. Meanwhile, damon_verify_ctx() is doing the damon_nr_regions() test. Because damon_verify_ctx() is called for each kdamond_call(), the test coverage from damon_verify_ctx() could be sufficient. Remove damon_nr_regions() verification.
--- mm/damon/core.c | 19 ------------------- 1 file changed, 19 deletions(-)
mm/damon/core.c +0 -19
@@ -686,27 +686,8 @@ 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 10/14 mm/damon/tests/core-kunit: add damon_set_regions() test cases
In reply to SeongJae Parkdamon_set_regions() is one of the main DAMON kernel API functions that set up the monitoring target memory region boundaries. Implement unit tests for verifying its basic functionalities.
--- mm/damon/tests/core-kunit.h | 142 ++++++++++++++++++++++++++++++------ 1 file changed, 120 insertions(+), 22 deletions(-)
mm/damon/tests/core-kunit.h +120 -22
@@ -390,41 +390,139 @@ static void damon_test_ops_registration(struct kunit *test) } } -static void damon_test_set_regions (struct kunit *test) +static void damon_test_set_regions_for (struct kunit *test, +struct damon_addr_range *old_ranges,int sz_old_ranges, +struct damon_addr_range *new_ranges,int sz_new_ranges, +unsigned long min_region_sz, +struct damon_addr_range *expect_ranges,int sz_expect_ranges) { -struct damon_target *t =damon_new_target (); -struct damon_region *r1, *r2; -struct damon_addr_range range = {.start =8 , .end =28 }; -unsigned long expects[] = {8 ,16 ,16 ,24 ,24 ,28 }; -int expect_idx =0 ; +struct damon_target *t;struct damon_region *r; +int i; + t =damon_new_target ();if (!t)kunit_skip (test,"target alloc fail" ); - r1 =damon_new_region (4 ,16 ); -if (!r1) { -damon_free_target (t); -kunit_skip (test,"region alloc fail" ); - } - r2 =damon_new_region (24 ,32 ); -if (!r2) { -damon_free_target (t); -damon_free_region (r1); -kunit_skip (test,"second region alloc fail" ); +for (i =0 ; i < sz_old_ranges; i++) { + r =damon_new_region (old_ranges[i].start, old_ranges[i].end); +if (!r) { +damon_destroy_target (t,NULL ); +kunit_skip (test,"%d-th r alloc fail\n" , i); + } +damon_add_region (r, t); } -damon_add_region (r1, t); -damon_add_region (r2, t); -damon_set_regions (t, &range,1 ,1 ); +damon_set_regions (t, new_ranges, sz_new_ranges, min_region_sz); -KUNIT_EXPECT_EQ (test,damon_nr_regions (t),3 ); +KUNIT_EXPECT_EQ (test,damon_nr_regions (t), sz_expect_ranges); +if (damon_nr_regions (t) != sz_expect_ranges) { +damon_destroy_target (t,NULL ); +return ; + } + i =0 ;damon_for_each_region (r, t) { -KUNIT_EXPECT_EQ (test, r->ar.start, expects[expect_idx++]); -KUNIT_EXPECT_EQ (test, r->ar.end, expects[expect_idx++]); +KUNIT_EXPECT_EQ (test, r->ar.start, expect_ranges[i].start); +KUNIT_EXPECT_EQ (test, r->ar.end, expect_ranges[i++].end); } +damon_destroy_target (t,NULL ); } +static void damon_test_set_regions (struct kunit *test) +{ +/* Initial build up on empty target. */ +damon_test_set_regions_for (test, + (struct damon_addr_range []){},0 , + (struct damon_addr_range []){ + {.start =5 , .end =15 }, + {.start =15 , .end =25 }, + },2 , +1 , + (struct damon_addr_range []){ + {.start =5 , .end =15 }, + {.start =15 , .end =25 }, + },2 ); +/* Un-intersecting regions should be removed. */ +damon_test_set_regions_for (test, + (struct damon_addr_range []){ + {.start =4 , .end =16 }, + {.start =24 , .end =32 }, + },2 , + (struct damon_addr_range []){ + {.start =18 , .end =23 }, + },1 , +1 , + (struct damon_addr_range []){ + {.start =18 , .end =23 }, + },1 ); +/* +* Holes should be filled up with new regions. +* +* old: [4, 16) [24, 32) +* new: [8, 28) +* expect: [8, 16)[16,24),[24, 28) +*/ +damon_test_set_regions_for (test, + (struct damon_addr_range []){ + {.start =4 , .end =16 }, + {.start =24 , .end =32 }, + },2 , + (struct damon_addr_range []){ + {.start =8 , .end =28 }, + },1 , +1 , + (struct damon_addr_range []){ + {.start =8 , .end =16 }, + {.start =16 , .end =24 }, + {.start =24 , .end =28 }, + },3 ); +/* +* New regions should be able to be appended. +* +* old: [0, 4)[4, 17) +* new: [0, 15) [25, 40) +* expect: [0, 4)[4, 15) [25, 40) +*/ +damon_test_set_regions_for (test, + (struct damon_addr_range []){ + {.start =0 , .end =4 }, + {.start =4 , .end =17 }, + },2 , + (struct damon_addr_range []){ + {.start =0 , .end =15 }, + {.start =25 , .end =40 }, + },2 , +1 , + (struct damon_addr_range []){ + {.start =0 , .end =4 }, + {.start =4 , .end =15 }, + {.start =25 , .end =40 }, + },3 ); +/* +* New regions should be able to be inserted. +* +* old: [0, 4) [42, 52) +* new: [0, 15) [25, 40) [44, 50) +* expect: [0, 15) [25, 40) [44, 50) +*/ +damon_test_set_regions_for (test, + (struct damon_addr_range []){ + {.start =0 , .end =4 }, + {.start =42 , .end =52 }, + },2 , + (struct damon_addr_range []){ + {.start =0 , .end =15 }, + {.start =25 , .end =40 }, + {.start =44 , .end =50 }, + },3 , +1 , + (struct damon_addr_range []){ + {.start =0 , .end =15 }, + {.start =25 , .end =40 }, + {.start =44 , .end =50 }, + },3 ); +} +static void damon_test_nr_accesses_to_accesses_bp (struct kunit *test) {struct damon_attrs attrs = {
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 11/14 selftests/damon/sysfs.py: stop kdamonds before failing
In reply to SeongJae ParkWhen an assertion is failed, sysfs.py DAMON selftest immediately exits the test program leaving the DAMON running behind. Many of the following tests need to start DAMON on their own. But because DAMON that was started by sysfs.py is still running, those start attempts fail, and the tests are failed or skipped. Update sysfs.py to stop DAMON before exiting the test program due to the assertion failure.
--- tools/testing/selftests/damon/sysfs.py | 4 ++++ 1 file changed, 4 insertions(+)
tools/testing/selftests/damon/sysfs.py +4 -0
@@ -24,9 +24,12 @@ def dump_damon_status_dict(pid): except Exception as e: return None, 'json.load fail (%s)' % e +kdamonds = None def fail(expectation, status): print('unexpected %s' % expectation) print(json.dumps(status, indent=4)) + if kdamonds is not None: + kdamonds.stop() exit(1) def assert_true(condition, expectation, status):
@@ -248,6 +251,7 @@ def assert_ctxs_committed(kdamonds): ctx.pause = False def main(): + global kdamonds kdamonds = _damon_sysfs.Kdamonds( [_damon_sysfs.Kdamond( contexts=[_damon_sysfs.DamonCtx(
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 12/14 selftests/damon/sysfs.sh: test monitoring intervals goal dir
In reply to SeongJae Parksysfs.sh DAMON selftest is not testing monitoring intervals goal directory. Add the test.
--- tools/testing/selftests/damon/sysfs.sh | 12 ++++++++++++ 1 file changed, 12 insertions(+)
tools/testing/selftests/damon/sysfs.sh +12 -0
@@ -282,6 +282,17 @@ test_targets() ensure_dir "$targets_dir/1" "not_exist" } + +test_intervals_goal() +{ + goal_dir=$1 + ensure_dir "$goal_dir" "exist" + ensure_file "$goal_dir/access_bp" "exist" "600" + ensure_file "$goal_dir/aggrs" "exist" "600" + ensure_file "$goal_dir/min_sample_us" "exist" "600" + ensure_file "$goal_dir/max_sample_us" "exist" "600" +} + test_intervals() { intervals_dir=$1
@@ -289,6 +300,7 @@ test_intervals() ensure_file "$intervals_dir/aggr_us" "exist" "600" ensure_file "$intervals_dir/sample_us" "exist" "600" ensure_file "$intervals_dir/update_us" "exist" "600" + test_intervals_goal "$intervals_dir/intervals_goal" } test_damon_filter()
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 13/14 selftests/damon/sysfs.sh: test addr_unit file existence
In reply to SeongJae Parksysfs.sh DAMON selftest is not testing the existence of addr_unit sysfs file. Add the test.
--- tools/testing/selftests/damon/sysfs.sh | 1 + 1 file changed, 1 insertion(+)
tools/testing/selftests/damon/sysfs.sh +1 -0
@@ -365,6 +365,7 @@ test_context() ensure_dir "$context_dir" "exist" ensure_file "$context_dir/avail_operations" "exit" 400 ensure_file "$context_dir/operations" "exist" 600 + ensure_file "$context_dir/addr_unit" "exist" 600 test_monitoring_attrs "$context_dir/monitoring_attrs" test_targets "$context_dir/targets" test_schemes "$context_dir/schemes"
Signature · 1 line
2.47.3
SeongJae Park
maintainerRFC PATCH 14/14 selftests/damon/sysfs.sh: test pause file existence
In reply to SeongJae Parksysfs.sh DAMON selftest is not testing the existence of the 'pause' sysfs file. Add the test.
--- tools/testing/selftests/damon/sysfs.sh | 1 + 1 file changed, 1 insertion(+)
tools/testing/selftests/damon/sysfs.sh +1 -0
@@ -366,6 +366,7 @@ test_context() ensure_file "$context_dir/avail_operations" "exit" 400 ensure_file "$context_dir/operations" "exist" 600 ensure_file "$context_dir/addr_unit" "exist" 600 + ensure_file "$context_dir/pause" "exist" 600 test_monitoring_attrs "$context_dir/monitoring_attrs" test_targets "$context_dir/targets" test_schemes "$context_dir/schemes"
Signature · 1 line
2.47.3