KVM: selftests: access_tracking_perf_test: skip the test when NUMA balancing is active
v2 cover KVM: selftests: access_tracking_perf_test: skip the test when NUMA balancing is active
Due to several issues which are unlikely to be fixed in the near future, the access_tracking_perf_test sanity check for how many pages are still clean after an iteration is not reliable when NUMA balancing is active.
This patch series refactors this test to skip this check by default automatically.
V2: adopted Sean's suggestions.
Best regards,
Maxim Levitsky
Maxim Levitsky (1):
KVM: selftests: access_tracking_perf_test: add option to skip the
sanity check
Sean Christopherson (1):
KVM: selftests: Extract guts of THP accessor to standalone sysfs
helpers
.../selftests/kvm/access_tracking_perf_test.c | 33 +++++++++++++-- .../testing/selftests/kvm/include/test_util.h | 1 + tools/testing/selftests/kvm/lib/test_util.c | 42 ++++++++++++++----- 3 files changed, 61 insertions(+), 15 deletions(-)
Signature · 1 line
2.26.3
PATCH v2 1/2 KVM: selftests: Extract guts of THP accessor to standalone sysfs helpers
From: Sean Christopherson <[email protected]>
Extract the guts of thp_configured() and get_trans_hugepagesz() to standalone helpers so that the core logic can be reused for other sysfs files, e.g. to query numa_balancing.
Opportunistically assert that the initial fscanf() read at least one byte, and add a comment explaining the second call to fscanf().
--- tools/testing/selftests/kvm/lib/test_util.c | 35 ++++++++++++++------- 1 file changed, 24 insertions(+), 11 deletions(-)
tools/testing/selftests/kvm/lib/test_util.c +24 -11
@@ -132,37 +132,50 @@ void print_skip(const char *fmt, ...)puts (", skipping test" ); } -bool thp_configured (void ) +static bool test_sysfs_path (const char *path) { -int ret;struct stat statbuf; +int ret; - ret =stat ("/sys/kernel/mm/transparent_hugepage" , &statbuf); + ret =stat (path, &statbuf);TEST_ASSERT (ret ==0 || (ret ==-1 && errno ==ENOENT ), -"Error in stating /sys/kernel/mm/transparent_hugepage" ); +"Error in stat()ing '%s'" , path);return ret ==0 ; } -size_t get_trans_hugepagesz (void ) +bool thp_configured (void ) +{ +return test_sysfs_path ("/sys/kernel/mm/transparent_hugepage" ); +} + +static size_t get_sysfs_val (const char *path) {size_t size;FILE *f;int ret; -TEST_ASSERT (thp_configured (),"THP is not configured in host kernel" ); - - f =fopen ("/sys/kernel/mm/transparent_hugepage/hpage_pmd_size" ,"r" ); -TEST_ASSERT (f !=NULL ,"Error in opening transparent_hugepage/hpage_pmd_size" ); + f =fopen (path,"r" ); +TEST_ASSERT (f,"Error opening '%s'" , path); ret =fscanf (f,"%ld" , &size); +TEST_ASSERT (ret >0 ,"Error reading '%s'" , path); + +/* Re-scan the input stream to verify the entire file was read. */ ret =fscanf (f,"%ld" , &size); -TEST_ASSERT (ret <1 ,"Error reading transparent_hugepage/hpage_pmd_size" ); -fclose (f); +TEST_ASSERT (ret <1 ,"Error reading '%s'" , path); +fclose (f);return size; } +size_t get_trans_hugepagesz (void ) +{ +TEST_ASSERT (thp_configured (),"THP is not configured in host kernel" ); + +return get_sysfs_val ("/sys/kernel/mm/transparent_hugepage/hpage_pmd_size" ); +} +size_t get_def_hugetlb_pagesz (void ) {char buf[64 ];
Signature · 1 line
2.26.3
PATCH v2 2/2 KVM: selftests: access_tracking_perf_test: add option to skip the sanity check
In reply to Maxim LevitskyAdd an option to skip sanity check of number of still idle pages, and set it by default to skip, in case hypervisor or NUMA balancing is detected.
--- .../selftests/kvm/access_tracking_perf_test.c | 33 ++++++++++++++++--- .../testing/selftests/kvm/include/test_util.h | 1 + tools/testing/selftests/kvm/lib/test_util.c | 7 ++++ 3 files changed, 37 insertions(+), 4 deletions(-)
tools/testing/selftests/kvm/access_tracking_perf_test.c +29 -4
@@ -65,6 +65,8 @@ static int vcpu_last_completed_iteration[KVM_MAX_VCPUS];/* Whether to overlap the regions of memory vCPUs access. */ static bool overlap_memory_access; +static int warn_on_too_many_idle_pages =-1 ; +struct test_params {/* The backing source for the region of memory. */ enum vm_mem_backing_src_type backing_src;
@@ -184,11 +186,10 @@ static void mark_vcpu_memory_idle(struct kvm_vm *vm,* are cached and the guest won't see the "idle" bit cleared. */ if (still_idle >= pages /10 ) { -#ifdef __x86_64__ -TEST_ASSERT (this_cpu_has (X86_FEATURE_HYPERVISOR ), +TEST_ASSERT (warn_on_too_many_idle_pages,"vCPU%d: Too many pages still idle (%lu out of %lu)" , vcpu_idx, still_idle, pages); -#endif +printf ("WARNING: vCPU%d: Too many pages still idle (%lu out of %lu), " "this will affect performance results.\n" , vcpu_idx, still_idle, pages);
@@ -342,6 +343,8 @@ static void help(char *name)printf (" -v: specify the number of vCPUs to run.\n" );printf (" -o: Overlap guest memory accesses instead of partitioning\n" " them into a separate region of memory for each vCPU.\n" ); +printf (" -w: Skip or force enable the check that after dirtying the guest memory, most (90%%) of \n" +"it is reported as dirty again (0/1)" );backing_src_help ("-s" );puts ("" );exit (0 );
@@ -359,7 +362,7 @@ int main(int argc, char *argv[])guest_modes_append_default (); -while ((opt =getopt (argc, argv,"hm:b:v:os:" )) !=-1 ) { +while ((opt =getopt (argc, argv,"hm:b:v:os:w:" )) !=-1 ) {switch (opt) {case 'm' :guest_modes_cmdline (optarg);
@@ -376,6 +379,11 @@ int main(int argc, char *argv[])case 's' : params.backing_src =parse_backing_src_type (optarg);break ; +case 'w' : + warn_on_too_many_idle_pages = +atoi_non_negative ("1 - enable warning, 0 - disable" , + optarg); +break ;case 'h' :default :help (argv[0 ]);
@@ -386,6 +394,23 @@ int main(int argc, char *argv[]) page_idle_fd =open ("/sys/kernel/mm/page_idle/bitmap" ,O_RDWR );__TEST_REQUIRE (page_idle_fd >=0 ,"CONFIG_IDLE_PAGE_TRACKING is not enabled" ); +if (warn_on_too_many_idle_pages ==-1 ) { +#ifdef __x86_64__ +if (this_cpu_has (X86_FEATURE_HYPERVISOR )) { +printf ("Skipping idle page count sanity check, because the test is run nested\n" ); + warn_on_too_many_idle_pages =0 ; + } else +#endif +if (is_numa_balancing_enabled ()) { +printf ("Skipping idle page count sanity check, because NUMA balance is enabled\n" ); + warn_on_too_many_idle_pages =0 ; + }else { + warn_on_too_many_idle_pages =1 ; + } + }else if (!warn_on_too_many_idle_pages) { +printf ("Skipping idle page count sanity check, because this was requested by the user\n" ); + } +close (page_idle_fd);for_each_guest_mode (run_test, ¶ms);
tools/testing/selftests/kvm/include/test_util.h +1 -0
@@ -153,6 +153,7 @@ bool is_backing_src_hugetlb(uint32_t i);void backing_src_help (const char *flag);enum vm_mem_backing_src_type parse_backing_src_type (const char *type_name);long get_run_delay (void ); +bool is_numa_balancing_enabled (void );/* * Whether or not the given source type is shared memory (as opposed to
tools/testing/selftests/kvm/lib/test_util.c +7 -0
@@ -176,6 +176,13 @@ size_t get_trans_hugepagesz(void)return get_sysfs_val ("/sys/kernel/mm/transparent_hugepage/hpage_pmd_size" ); } +bool is_numa_balancing_enabled (void ) +{ +if (!test_sysfs_path ("/proc/sys/kernel/numa_balancing" )) +return false; +return get_sysfs_val ("/proc/sys/kernel/numa_balancing" ) ==1 ; +} +size_t get_def_hugetlb_pagesz (void ) {char buf[64 ];
Signature · 1 line
2.26.3
Re: [PATCH v2 2/2] KVM: selftests: access_tracking_perf_test: add option to skip the sanity check
Quoted 6 lines · Maxim Levitsky
Add an option to skip sanity check of number of still idle pages, and set it by default to skip, in case hypervisor or NUMA balancing is detected.
Signed-off-by Maxim Levitsky <[email protected]>
Thanks Maxim! I'm still working on a respin of this test with MGLRU integration, like [1]. Sorry it's taking me so long. I'll apply my changes on top of yours.
[1]: https://lore.kernel.org/kvm/[email protected]/
Quoted 20 lines
--- .../selftests/kvm/access_tracking_perf_test.c | 33 ++++++++++++++++--- .../testing/selftests/kvm/include/test_util.h | 1 + tools/testing/selftests/kvm/lib/test_util.c | 7 ++++ 3 files changed, 37 insertions(+), 4 deletions(-)tools/testing/selftests/kvm/access_tracking_perf_test.c +2
-0@@ -65,6 +65,8 @@ static int vcpu_last_completed_iteration[KVM_MAX_VCPUS];/* Whether to overlap the regions of memory vCPUs access. */ static bool overlap_memory_access; +static int warn_on_too_many_idle_pages =-1 ; +struct test_params {/* The backing source for the region of memory. */ enum vm_mem_backing_src_type backing_src;
tools/testing/selftests/kvm/access_tracking_perf_test.c static void mark_vcpu_memory_idle(struct kvm_vm *vm, · line 189
Quoted 6 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +1
-2@@ -184,11 +186,10 @@ static void mark_vcpu_memory_idle(struct kvm_vm *vm,* are cached and the guest won't see the "idle" bit cleared. */ if (still_idle >= pages /10 ) { -#ifdef __x86_64__ - TEST_ASSERT(this_cpu_has(X86_FEATURE_HYPERVISOR), + TEST_ASSERT(warn_on_too_many_idle_pages,
I think this assertion is flipped (or how warn_on_too_many_idle_pages is being set is flipped, see below).
Quoted 26 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +4
-2"vCPU%d: Too many pages still idle (%lu out of %lu)" , vcpu_idx, still_idle, pages); -#endif +printf ("WARNING: vCPU%d: Too many pages still idle (%lu out of %lu), " "this will affect performance results.\n" , vcpu_idx, still_idle, pages);@@ -342,6 +343,8 @@ static void help(char *name)printf (" -v: specify the number of vCPUs to run.\n" );printf (" -o: Overlap guest memory accesses instead of partitioning\n" " them into a separate region of memory for each vCPU.\n" ); +printf (" -w: Skip or force enable the check that after dirtying the guest memory, most (90%%) of \n" +"it is reported as dirty again (0/1)" );backing_src_help ("-s" );puts ("" );exit (0 );@@ -359,7 +362,7 @@ int main(int argc, char *argv[])guest_modes_append_default (); -while ((opt =getopt (argc, argv,"hm:b:v:os:" )) !=-1 ) { +while ((opt =getopt (argc, argv,"hm:b:v:os:w:" )) !=-1 ) {switch (opt) {case 'm' :guest_modes_cmdline (optarg);
tools/testing/selftests/kvm/access_tracking_perf_test.c int main(int argc, char *argv[]) · line 385
Quoted 7 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +4
-0@@ -376,6 +379,11 @@ int main(int argc, char *argv[])case 's' : params.backing_src =parse_backing_src_type (optarg);break ; +case 'w' : + warn_on_too_many_idle_pages = +atoi_non_negative ("1 - enable warning, 0 - disable" , + optarg);
We still get a "warning" either way, right? Maybe this should be called "fail_on_too_many_idle_pages" (in which case the above assertion is indeed flipped). Or "warn_on_too_many_idle_pages" should mean *only* warn, i.e., *don't* fail, in which case, below we need to flip how we set it below.
Quoted 61 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +18
-0+break ;case 'h' :default :help (argv[0 ]);@@ -386,6 +394,23 @@ int main(int argc, char *argv[]) page_idle_fd =open ("/sys/kernel/mm/page_idle/bitmap" ,O_RDWR );__TEST_REQUIRE (page_idle_fd >=0 ,"CONFIG_IDLE_PAGE_TRACKING is not enabled" ); +if (warn_on_too_many_idle_pages ==-1 ) { +#ifdef __x86_64__ +if (this_cpu_has (X86_FEATURE_HYPERVISOR )) { +printf ("Skipping idle page count sanity check, because the test is run nested\n" ); + warn_on_too_many_idle_pages =0 ; + } else +#endif +if (is_numa_balancing_enabled ()) { +printf ("Skipping idle page count sanity check, because NUMA balance is enabled\n" ); + warn_on_too_many_idle_pages =0 ; + }else { + warn_on_too_many_idle_pages =1 ; + } + }else if (!warn_on_too_many_idle_pages) { +printf ("Skipping idle page count sanity check, because this was requested by the user\n" ); + } +close (page_idle_fd);for_each_guest_mode (run_test, ¶ms);tools/testing/selftests/kvm/include/test_util.h +1
-0@@ -153,6 +153,7 @@ bool is_backing_src_hugetlb(uint32_t i);void backing_src_help (const char *flag);enum vm_mem_backing_src_type parse_backing_src_type (const char *type_name);long get_run_delay (void ); +bool is_numa_balancing_enabled (void );/* * Whether or not the given source type is shared memory (as opposed to tools/testing/selftests/kvm/lib/test_util.c +7
-1@@ -176,6 +176,13 @@ size_t get_trans_hugepagesz(void)return get_sysfs_val ("/sys/kernel/mm/transparent_hugepage/hpage_pmd_size" ); } +bool is_numa_balancing_enabled (void ) +{ +if (!test_sysfs_path ("/proc/sys/kernel/numa_balancing" )) +return false; +return get_sysfs_val ("/proc/sys/kernel/numa_balancing" ) ==1 ; +} +size_t get_def_hugetlb_pagesz (void ) {char buf[64 ]; --2.26.3
Sean Christopherson
reviewerRe: [PATCH v2 2/2] KVM: selftests: access_tracking_perf_test: add option to skip the sanity check
Quoted 75 lines · James Houghton
Quoted 6 lines · Maxim Levitsky
Add an option to skip sanity check of number of still idle pages, and set it by default to skip, in case hypervisor or NUMA balancing is detected.
Signed-off-by Maxim Levitsky <[email protected]>Thanks Maxim! I'm still working on a respin of this test with MGLRU integration, like [1]. Sorry it's taking me so long. I'll apply my changes on top of yours.
[1]: https://lore.kernel.org/kvm/[email protected]/
Quoted 20 lines
--- .../selftests/kvm/access_tracking_perf_test.c | 33 ++++++++++++++++--- .../testing/selftests/kvm/include/test_util.h | 1 + tools/testing/selftests/kvm/lib/test_util.c | 7 ++++ 3 files changed, 37 insertions(+), 4 deletions(-)tools/testing/selftests/kvm/access_tracking_perf_test.c +2
-0@@ -65,6 +65,8 @@ static int vcpu_last_completed_iteration[KVM_MAX_VCPUS];/* Whether to overlap the regions of memory vCPUs access. */ static bool overlap_memory_access; +static int warn_on_too_many_idle_pages =-1 ; +struct test_params {/* The backing source for the region of memory. */ enum vm_mem_backing_src_type backing_src;tools/testing/selftests/kvm/access_tracking_perf_test.c static void mark_vcpu_memory_idle(struct kvm_vm *vm, · line 189
Quoted 6 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +1
-2@@ -184,11 +186,10 @@ static void mark_vcpu_memory_idle(struct kvm_vm *vm,* are cached and the guest won't see the "idle" bit cleared. */ if (still_idle >= pages /10 ) { -#ifdef __x86_64__ - TEST_ASSERT(this_cpu_has(X86_FEATURE_HYPERVISOR), + TEST_ASSERT(warn_on_too_many_idle_pages,I think this assertion is flipped (or how
warn_on_too_many_idle_pagesis being set is flipped, see below).Quoted 31 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +6
-2"vCPU%d: Too many pages still idle (%lu out of %lu)" , vcpu_idx, still_idle, pages); -#endif +printf ("WARNING: vCPU%d: Too many pages still idle (%lu out of %lu), " "this will affect performance results.\n" , vcpu_idx, still_idle, pages);@@ -342,6 +343,8 @@ static void help(char *name)printf (" -v: specify the number of vCPUs to run.\n" );printf (" -o: Overlap guest memory accesses instead of partitioning\n" " them into a separate region of memory for each vCPU.\n" ); +printf (" -w: Skip or force enable the check that after dirtying the guest memory, most (90%%) of \n" +"it is reported as dirty again (0/1)" );backing_src_help ("-s" );puts ("" );exit (0 );@@ -359,7 +362,7 @@ int main(int argc, char *argv[])guest_modes_append_default (); -while ((opt =getopt (argc, argv,"hm:b:v:os:" )) !=-1 ) { +while ((opt =getopt (argc, argv,"hm:b:v:os:w:" )) !=-1 ) {switch (opt) {case 'm' :guest_modes_cmdline (optarg);@@ -376,6 +379,11 @@ int main(int argc, char *argv[])case 's' : params.backing_src =parse_backing_src_type (optarg);break ; +case 'w' : + warn_on_too_many_idle_pages =
Quoted 8 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c int main(int argc, char *argv[]) · line 385
Quoted 2 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +2
-0+atoi_non_negative ("1 - enable warning, 0 - disable" , + optarg);We still get a "warning" either way, right? Maybe this should be called "
fail_on_too_many_idle_pages" (in which case the above assertion is indeed flipped). Or "warn_on_too_many_idle_pages" should mean *only* warn, i.e., *don't* fail, in which case, below we need to flip how we set it below.
Agreed. I like the "warn" terminology, Maybe this?
printf(" -w: Control whether the test warns or fails if more than 10%\n"
" of pages are still seen as idle/old after accessing guest\n"
" memory. >0 == warn only, 0 == fail, <0 == auto. For auto\n"
" mode, the test fails by default, but switches to warn only\n"
" if NUMA balancing is enabled or the test detects it's running\n"
" in a VM.");
And let the user explicitly select auto:
case 'w': warn_only = atoi_paranoid(optarg); break;
Then the auto resolving works as below, and as James pointed out, the assert becomes
TEST_ASSERT(!warn_only, ....);
Quoted 16 lines
Quoted 16 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +8
-0+break ;case 'h' :default :help (argv[0 ]);@@ -386,6 +394,23 @@ int main(int argc, char *argv[]) page_idle_fd =open ("/sys/kernel/mm/page_idle/bitmap" ,O_RDWR );__TEST_REQUIRE (page_idle_fd >=0 ,"CONFIG_IDLE_PAGE_TRACKING is not enabled" ); +if (warn_on_too_many_idle_pages ==-1 ) { +#ifdef __x86_64__ +if (this_cpu_has (X86_FEATURE_HYPERVISOR )) { +printf ("Skipping idle page count sanity check, because the test is run nested\n" ); + warn_on_too_many_idle_pages =0 ; + } else +#endif
Quoted 8 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c int main(int argc, char *argv[]) · line 411
Quoted 8 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +8
-0+if (is_numa_balancing_enabled ()) { +printf ("Skipping idle page count sanity check, because NUMA balance is enabled\n" ); + warn_on_too_many_idle_pages =0 ; + }else { + warn_on_too_many_idle_pages =1 ; + } + }else if (!warn_on_too_many_idle_pages) { +printf ("Skipping idle page count sanity check, because this was requested by the user\n" );
Eh, I vote to omit this. The sanity check is still there, it's just degraded to a warn. I'm not totally against it, just seems superfluous and potentially confusing.
Re: [PATCH v2 2/2] KVM: selftests: access_tracking_perf_test: add option to skip the sanity check
Quoted 87 lines · Sean Christopherson
Quoted 75 lines · James Houghton
Quoted 6 lines · Maxim Levitsky
Add an option to skip sanity check of number of still idle pages, and set it by default to skip, in case hypervisor or NUMA balancing is detected.
Signed-off-by Maxim Levitsky <[email protected]>Thanks Maxim! I'm still working on a respin of this test with MGLRU integration, like [1]. Sorry it's taking me so long. I'll apply my changes on top of yours.
[1]: https://lore.kernel.org/kvm/[email protected]/
Quoted 20 lines
--- .../selftests/kvm/access_tracking_perf_test.c | 33 ++++++++++++++++--- .../testing/selftests/kvm/include/test_util.h | 1 + tools/testing/selftests/kvm/lib/test_util.c | 7 ++++ 3 files changed, 37 insertions(+), 4 deletions(-)tools/testing/selftests/kvm/access_tracking_perf_test.c +2
-0@@ -65,6 +65,8 @@ static int vcpu_last_completed_iteration[KVM_MAX_VCPUS];/* Whether to overlap the regions of memory vCPUs access. */ static bool overlap_memory_access; +static int warn_on_too_many_idle_pages =-1 ; +struct test_params {/* The backing source for the region of memory. */ enum vm_mem_backing_src_type backing_src;tools/testing/selftests/kvm/access_tracking_perf_test.c static void mark_vcpu_memory_idle(struct kvm_vm *vm, · line 189
Quoted 6 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +1
-2@@ -184,11 +186,10 @@ static void mark_vcpu_memory_idle(struct kvm_vm *vm,* are cached and the guest won't see the "idle" bit cleared. */ if (still_idle >= pages /10 ) { -#ifdef __x86_64__ - TEST_ASSERT(this_cpu_has(X86_FEATURE_HYPERVISOR), + TEST_ASSERT(warn_on_too_many_idle_pages,I think this assertion is flipped (or how
warn_on_too_many_idle_pagesis being set is flipped, see below).Quoted 31 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +6
-2"vCPU%d: Too many pages still idle (%lu out of %lu)" , vcpu_idx, still_idle, pages); -#endif +printf ("WARNING: vCPU%d: Too many pages still idle (%lu out of %lu), " "this will affect performance results.\n" , vcpu_idx, still_idle, pages);@@ -342,6 +343,8 @@ static void help(char *name)printf (" -v: specify the number of vCPUs to run.\n" );printf (" -o: Overlap guest memory accesses instead of partitioning\n" " them into a separate region of memory for each vCPU.\n" ); +printf (" -w: Skip or force enable the check that after dirtying the guest memory, most (90%%) of \n" +"it is reported as dirty again (0/1)" );backing_src_help ("-s" );puts ("" );exit (0 );@@ -359,7 +362,7 @@ int main(int argc, char *argv[])guest_modes_append_default (); -while ((opt =getopt (argc, argv,"hm:b:v:os:" )) !=-1 ) { +while ((opt =getopt (argc, argv,"hm:b:v:os:w:" )) !=-1 ) {switch (opt) {case 'm' :guest_modes_cmdline (optarg);@@ -376,6 +379,11 @@ int main(int argc, char *argv[])case 's' : params.backing_src =parse_backing_src_type (optarg);break ; +case 'w' : + warn_on_too_many_idle_pages =Quoted 8 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c int main(int argc, char *argv[]) · line 385
Quoted 2 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +2
-0+atoi_non_negative ("1 - enable warning, 0 - disable" , + optarg);We still get a "warning" either way, right? Maybe this should be called "
fail_on_too_many_idle_pages" (in which case the above assertion is indeed flipped). Or "warn_on_too_many_idle_pages" should mean *only* warn, i.e., *don't* fail, in which case, below we need to flip how we set it below.
Quoted 8 lines
Agreed. I like the "warn" terminology, Maybe this?
printf(" -w: Control whether the test warns or fails if more than 10%\n"
" of pages are still seen as idle/old after accessing guest\n"
" memory. >0 == warn only, 0 == fail, <0 == auto. For auto\n"
" mode, the test fails by default, but switches to warn only\n"
" if NUMA balancing is enabled or the test detects it's running\n"
" in a VM.");
LGTM.
Quoted 11 lines
And let the user explicitly select auto:
case 'w': warn_only = atoi_paranoid(optarg); break;Then the auto resolving works as below, and as James pointed out, the assert becomes
TEST_ASSERT(!warn_only, ....);
I think the auto-resolving below needs to be flipped, and the TEST_ASSERT should be for warn_only, not !warn_only.
If warn_only == 1, the assert should pass.
Quoted 20 lines
Quoted 20 lines
Quoted 20 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +11
-0+break ;case 'h' :default :help (argv[0 ]);@@ -386,6 +394,23 @@ int main(int argc, char *argv[]) page_idle_fd =open ("/sys/kernel/mm/page_idle/bitmap" ,O_RDWR );__TEST_REQUIRE (page_idle_fd >=0 ,"CONFIG_IDLE_PAGE_TRACKING is not enabled" ); +if (warn_on_too_many_idle_pages ==-1 ) { +#ifdef __x86_64__ +if (this_cpu_has (X86_FEATURE_HYPERVISOR )) { +printf ("Skipping idle page count sanity check, because the test is run nested\n" ); + warn_on_too_many_idle_pages =0 ; + } else +#endif +if (is_numa_balancing_enabled ()) { +printf ("Skipping idle page count sanity check, because NUMA balance is enabled\n" ); + warn_on_too_many_idle_pages =0 ;
Quoted 8 lines
Quoted 5 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c int main(int argc, char *argv[]) · line 411
Quoted 5 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +5
-0+ }else { + warn_on_too_many_idle_pages =1 ; + } + }else if (!warn_on_too_many_idle_pages) { +printf ("Skipping idle page count sanity check, because this was requested by the user\n" );Eh, I vote to omit this. The sanity check is still there, it's just degraded to a warn. I'm not totally against it, just seems superfluous and potentially confusing.
I agree, it's not adding much.
Separately: I've finished the MGLRU version of this test. It uses MGLRU if it is available, and marking pages as idle is much faster when using it. If MGLRU is enabled but otherwise not usable, the test fails, as the idle page bitmap is no longer usable for this test.
I'm happy to post a new version of Maxim's patch with the MGLRU patches too, Maxim, if you're okay with that.
Sean Christopherson
reviewerRe: [PATCH v2 2/2] KVM: selftests: access_tracking_perf_test: add option to skip the sanity check
Quoted 10 lines · James Houghton
Quoted 4 lines · Sean Christopherson
Then the auto resolving works as below, and as James pointed out, the assert becomes
TEST_ASSERT(!warn_only, ....);I think the auto-resolving below needs to be flipped, and the
TEST_ASSERTshould be forwarn_only, not!warn_only.If
warn_only== 1, the assert should pass.
/facepalm, yep
Quoted 36 lines
Quoted 18 lines
Quoted 18 lines
Quoted 18 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +11
-0+break ;case 'h' :default :help (argv[0 ]);@@ -386,6 +394,23 @@ int main(int argc, char *argv[]) page_idle_fd =open ("/sys/kernel/mm/page_idle/bitmap" ,O_RDWR );__TEST_REQUIRE (page_idle_fd >=0 ,"CONFIG_IDLE_PAGE_TRACKING is not enabled" ); +if (warn_on_too_many_idle_pages ==-1 ) { +#ifdef __x86_64__ +if (this_cpu_has (X86_FEATURE_HYPERVISOR )) { +printf ("Skipping idle page count sanity check, because the test is run nested\n" ); + warn_on_too_many_idle_pages =0 ; + } else +#endif +if (is_numa_balancing_enabled ()) { +printf ("Skipping idle page count sanity check, because NUMA balance is enabled\n" ); + warn_on_too_many_idle_pages =0 ;Quoted 8 lines
Quoted 5 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c int main(int argc, char *argv[]) · line 411
Quoted 5 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +5
-0+ }else { + warn_on_too_many_idle_pages =1 ; + } + }else if (!warn_on_too_many_idle_pages) { +printf ("Skipping idle page count sanity check, because this was requested by the user\n" );Eh, I vote to omit this. The sanity check is still there, it's just degraded to a warn. I'm not totally against it, just seems superfluous and potentially confusing.
I agree, it's not adding much.
Separately: I've finished the MGLRU version of this test. It uses MGLRU if it is available, and marking pages as idle is much faster when using it. If MGLRU is enabled but otherwise not usable, the test fails, as the idle page bitmap is no longer usable for this test.
I'm happy to post a new version of Maxim's patch with the MGLRU patches too, Maxim, if you're okay with that.
Re: [PATCH v2 2/2] KVM: selftests: access_tracking_perf_test: add option to skip the sanity check
In reply to James HoughtonQuoted 27 lines · James Houghton
Quoted 5 lines · Maxim Levitsky
Add an option to skip sanity check of number of still idle pages, and set it by default to skip, in case hypervisor or NUMA balancing is detected.
Signed-off-by Maxim Levitsky <[email protected]>Thanks Maxim! I'm still working on a respin of this test with MGLRU integration, like [1]. Sorry it's taking me so long. I'll apply my changes on top of yours.
[1]: https://lore.kernel.org/kvm/[email protected]/
Quoted 14 lines
--- .../selftests/kvm/access_tracking_perf_test.c | 33 ++++++++++++++++--- .../testing/selftests/kvm/include/test_util.h | 1 + tools/testing/selftests/kvm/lib/test_util.c | 7 ++++ 3 files changed, 37 insertions(+), 4 deletions(-)tools/testing/selftests/kvm/access_tracking_perf_test.c +0
-0@@ -65,6 +65,8 @@ static int vcpu_last_completed_iteration[KVM_MAX_VCPUS];/* Whether to overlap the regions of memory vCPUs access. */ static bool overlap_memory_access;
Quoted 15 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c static void mark_vcpu_memory_idle(struct kvm_vm *vm, · line 189
Quoted 12 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +3
-2+static int warn_on_too_many_idle_pages =-1 ; +struct test_params {/* The backing source for the region of memory. */ enum vm_mem_backing_src_type backing_src;@@ -184,11 +186,10 @@ static void mark_vcpu_memory_idle(struct kvm_vm *vm,* are cached and the guest won't see the "idle" bit cleared. */ if (still_idle >= pages /10 ) { -#ifdef __x86_64__ - TEST_ASSERT(this_cpu_has(X86_FEATURE_HYPERVISOR), + TEST_ASSERT(warn_on_too_many_idle_pages,I think this assertion is flipped (or how
warn_on_too_many_idle_pagesis being set is flipped, see below).
Yes it is no doubt about.
I didn't notice this when I flipped the meaning
of the variable as Sean suggested.
Thanks!
Best regards,
Maxim Levitsky
Quoted 102 lines
Quoted 27 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +4
-2"vCPU%d: Too many pages still idle (%lu out of %lu)" , vcpu_idx, still_idle, pages); -#endif +printf ("WARNING: vCPU%d: Too many pages still idle (%lu out of %lu), " "this will affect performance results.\n" , vcpu_idx, still_idle, pages);@@ -342,6 +343,8 @@ static void help(char *name)printf (" -v: specify the number of vCPUs to run.\n" );printf (" -o: Overlap guest memory accesses instead of partitioning\n" " them into a separate region of memory for each vCPU.\n" ); +printf (" -w: Skip or force enable the check that after dirtying the guest memory, most (90%%) of \n" +"it is reported as dirty again (0/1)" );backing_src_help ("-s" );puts ("" );exit (0 );@@ -359,7 +362,7 @@ int main(int argc, char *argv[])guest_modes_append_default (); -while ((opt =getopt (argc, argv,"hm:b:v:os:" )) !=-1 ) { +while ((opt =getopt (argc, argv,"hm:b:v:os:w:" )) !=-1 ) {switch (opt) {case 'm' :guest_modes_cmdline (optarg);tools/testing/selftests/kvm/access_tracking_perf_test.c int main(int argc, char *argv[]) · line 385
Quoted 7 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +4
-0@@ -376,6 +379,11 @@ int main(int argc, char *argv[])case 's' : params.backing_src =parse_backing_src_type (optarg);break ; +case 'w' : + warn_on_too_many_idle_pages = +atoi_non_negative ("1 - enable warning, 0 - disable" , + optarg);We still get a "warning" either way, right? Maybe this should be called "
fail_on_too_many_idle_pages" (in which case the above assertion is indeed flipped). Or "warn_on_too_many_idle_pages" should mean *only* warn, i.e., *don't* fail, in which case, below we need to flip how we set it below.Quoted 61 lines
tools/testing/selftests/kvm/access_tracking_perf_test.c +18
-0+break ;case 'h' :default :help (argv[0 ]);@@ -386,6 +394,23 @@ int main(int argc, char *argv[]) page_idle_fd =open ("/sys/kernel/mm/page_idle/bitmap" ,O_RDWR );__TEST_REQUIRE (page_idle_fd >=0 ,"CONFIG_IDLE_PAGE_TRACKING is not enabled" ); +if (warn_on_too_many_idle_pages ==-1 ) { +#ifdef __x86_64__ +if (this_cpu_has (X86_FEATURE_HYPERVISOR )) { +printf ("Skipping idle page count sanity check, because the test is run nested\n" ); + warn_on_too_many_idle_pages =0 ; + } else +#endif +if (is_numa_balancing_enabled ()) { +printf ("Skipping idle page count sanity check, because NUMA balance is enabled\n" ); + warn_on_too_many_idle_pages =0 ; + }else { + warn_on_too_many_idle_pages =1 ; + } + }else if (!warn_on_too_many_idle_pages) { +printf ("Skipping idle page count sanity check, because this was requested by the user\n" ); + } +close (page_idle_fd);for_each_guest_mode (run_test, ¶ms);tools/testing/selftests/kvm/include/test_util.h +1
-0@@ -153,6 +153,7 @@ bool is_backing_src_hugetlb(uint32_t i);void backing_src_help (const char *flag);enum vm_mem_backing_src_type parse_backing_src_type (const char *type_name);long get_run_delay (void ); +bool is_numa_balancing_enabled (void );/* * Whether or not the given source type is shared memory (as opposed to tools/testing/selftests/kvm/lib/test_util.c +7
-1@@ -176,6 +176,13 @@ size_t get_trans_hugepagesz(void)return get_sysfs_val ("/sys/kernel/mm/transparent_hugepage/hpage_pmd_size" ); } +bool is_numa_balancing_enabled (void ) +{ +if (!test_sysfs_path ("/proc/sys/kernel/numa_balancing" )) +return false; +return get_sysfs_val ("/proc/sys/kernel/numa_balancing" ) ==1 ; +} +size_t get_def_hugetlb_pagesz (void ) {char buf[64 ]; --2.26.3