* [PATCH 0/2] tests: migration-test: Fix agressive test skip, add dirty ring test @ 2021-06-10 0:52 Peter Xu 2021-06-10 0:52 ` [PATCH 1/2] tests: migration-test: Still run the rest even if uffd missing Peter Xu 2021-06-10 0:52 ` [PATCH 2/2] tests: migration-test: Add dirty ring test Peter Xu 0 siblings, 2 replies; 7+ messages in thread From: Peter Xu @ 2021-06-10 0:52 UTC (permalink / raw) To: qemu-devel Cc: Paolo Bonzini, Leonardo Bras Soares Passos, Dr . David Alan Gilbert, peterx, Juan Quintela Based-on: <20210609014355.217110-1-peterx@redhat.com> Patch 1 is a fix for migration test not really running on new kernels. The problem is uffd check now will constantly fail after upstream commit 37cd0575b8510159 - that means any host kernel newer than 5.11. Patch 1 makes it slightly better by only skipping the two postcopy tests that needs uffd on these kernels. When we want to run the full test, we can do: $ sudo QTEST_QEMU_BINARY=./qemu-system-x86_64 ./tests/qtest/migration-test Then the uffd check will pass, and postcopy tests will be run. Patch 2 of this series adds the dirty ring test that just got merged into qemu. It needs the other patch "[PATCH] KVM: Fix dirty ring mmap incorrect size due to renaming accident", and that's majorly why we need the "Based-on" tag. Not sure what's the easiest way for the series as it'll depend on the other kvm patch. Perhaps if I can try to get ack from Dave so Paolo could queue it too along with the kvm fix (for either the whole series or patch 2 only)? I'll leave that to maintainers to decide.. Please review, thanks. Peter Xu (2): tests: migration-test: Still run the rest even if uffd missing tests: migration-test: Add dirty ring test tests/qtest/migration-test.c | 62 ++++++++++++++++++++++++++++++------ 1 file changed, 53 insertions(+), 9 deletions(-) -- 2.31.1 ^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 1/2] tests: migration-test: Still run the rest even if uffd missing 2021-06-10 0:52 [PATCH 0/2] tests: migration-test: Fix agressive test skip, add dirty ring test Peter Xu @ 2021-06-10 0:52 ` Peter Xu 2021-06-10 18:56 ` Dr. David Alan Gilbert 2021-06-10 0:52 ` [PATCH 2/2] tests: migration-test: Add dirty ring test Peter Xu 1 sibling, 1 reply; 7+ messages in thread From: Peter Xu @ 2021-06-10 0:52 UTC (permalink / raw) To: qemu-devel Cc: Paolo Bonzini, Leonardo Bras Soares Passos, Dr . David Alan Gilbert, peterx, Juan Quintela Currently we'll skip the whole migration-test if uffd missing. It's a bit harsh - we can still run the rest besides postcopy! Enable them when we still can. It'll happen more frequently now after kernel UFFD_USER_MODE_ONLY introduced in commit 37cd0575b8510159, as qemu test normally requires kernel faults. One alternative is we disable kvm and create the uffd with UFFD_USER_MODE_ONLY for all postcopy tests, however to be simple for now just skip postcopy tests only by default. If we wanna run them use "sudo" or root, they'll still work. In all cases, it's still better than running nothing for migration-test. Signed-off-by: Peter Xu <peterx@redhat.com> --- tests/qtest/migration-test.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c index 2b028df6875..d9225f58d4d 100644 --- a/tests/qtest/migration-test.c +++ b/tests/qtest/migration-test.c @@ -1376,10 +1376,6 @@ int main(int argc, char **argv) g_test_init(&argc, &argv, NULL); - if (!ufd_version_check()) { - return g_test_run(); - } - /* * On ppc64, the test only works with kvm-hv, but not with kvm-pr and TCG * is touchy due to race conditions on dirty bits (especially on PPC for @@ -1416,8 +1412,11 @@ int main(int argc, char **argv) module_call_init(MODULE_INIT_QOM); - qtest_add_func("/migration/postcopy/unix", test_postcopy); - qtest_add_func("/migration/postcopy/recovery", test_postcopy_recovery); + if (ufd_version_check()) { + qtest_add_func("/migration/postcopy/unix", test_postcopy); + qtest_add_func("/migration/postcopy/recovery", test_postcopy_recovery); + } + qtest_add_func("/migration/bad_dest", test_baddest); qtest_add_func("/migration/precopy/unix", test_precopy_unix); qtest_add_func("/migration/precopy/tcp", test_precopy_tcp); -- 2.31.1 ^ permalink raw reply related [flat|nested] 7+ messages in thread
* Re: [PATCH 1/2] tests: migration-test: Still run the rest even if uffd missing 2021-06-10 0:52 ` [PATCH 1/2] tests: migration-test: Still run the rest even if uffd missing Peter Xu @ 2021-06-10 18:56 ` Dr. David Alan Gilbert 0 siblings, 0 replies; 7+ messages in thread From: Dr. David Alan Gilbert @ 2021-06-10 18:56 UTC (permalink / raw) To: Peter Xu Cc: Paolo Bonzini, Leonardo Bras Soares Passos, qemu-devel, Juan Quintela * Peter Xu (peterx@redhat.com) wrote: > Currently we'll skip the whole migration-test if uffd missing. > > It's a bit harsh - we can still run the rest besides postcopy! Enable them > when we still can. > > It'll happen more frequently now after kernel UFFD_USER_MODE_ONLY introduced in > commit 37cd0575b8510159, as qemu test normally requires kernel faults. One > alternative is we disable kvm and create the uffd with UFFD_USER_MODE_ONLY for > all postcopy tests, however to be simple for now just skip postcopy tests only > by default. If we wanna run them use "sudo" or root, they'll still work. In > all cases, it's still better than running nothing for migration-test. > > Signed-off-by: Peter Xu <peterx@redhat.com> Ouch! Yes; that check was originally in a standalone test file for postcopy that's then morphed into the full test over a few years. Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com> > --- > tests/qtest/migration-test.c | 11 +++++------ > 1 file changed, 5 insertions(+), 6 deletions(-) > > diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c > index 2b028df6875..d9225f58d4d 100644 > --- a/tests/qtest/migration-test.c > +++ b/tests/qtest/migration-test.c > @@ -1376,10 +1376,6 @@ int main(int argc, char **argv) > > g_test_init(&argc, &argv, NULL); > > - if (!ufd_version_check()) { > - return g_test_run(); > - } > - > /* > * On ppc64, the test only works with kvm-hv, but not with kvm-pr and TCG > * is touchy due to race conditions on dirty bits (especially on PPC for > @@ -1416,8 +1412,11 @@ int main(int argc, char **argv) > > module_call_init(MODULE_INIT_QOM); > > - qtest_add_func("/migration/postcopy/unix", test_postcopy); > - qtest_add_func("/migration/postcopy/recovery", test_postcopy_recovery); > + if (ufd_version_check()) { > + qtest_add_func("/migration/postcopy/unix", test_postcopy); > + qtest_add_func("/migration/postcopy/recovery", test_postcopy_recovery); > + } > + > qtest_add_func("/migration/bad_dest", test_baddest); > qtest_add_func("/migration/precopy/unix", test_precopy_unix); > qtest_add_func("/migration/precopy/tcp", test_precopy_tcp); > -- > 2.31.1 > -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK ^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 2/2] tests: migration-test: Add dirty ring test 2021-06-10 0:52 [PATCH 0/2] tests: migration-test: Fix agressive test skip, add dirty ring test Peter Xu 2021-06-10 0:52 ` [PATCH 1/2] tests: migration-test: Still run the rest even if uffd missing Peter Xu @ 2021-06-10 0:52 ` Peter Xu 2021-06-10 19:01 ` Dr. David Alan Gilbert 1 sibling, 1 reply; 7+ messages in thread From: Peter Xu @ 2021-06-10 0:52 UTC (permalink / raw) To: qemu-devel Cc: Paolo Bonzini, Leonardo Bras Soares Passos, Dr . David Alan Gilbert, peterx, Juan Quintela Add dirty ring test if kernel supports it. Add the dirty ring parameter on source should be mostly enough, but let's change the dest too to make them match always. Signed-off-by: Peter Xu <peterx@redhat.com> --- tests/qtest/migration-test.c | 51 +++++++++++++++++++++++++++++++++--- 1 file changed, 48 insertions(+), 3 deletions(-) diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c index d9225f58d4d..cc6e396d1a2 100644 --- a/tests/qtest/migration-test.c +++ b/tests/qtest/migration-test.c @@ -12,6 +12,7 @@ #include "qemu/osdep.h" +#include <linux/kvm.h> #include "libqos/libqtest.h" #include "qapi/error.h" #include "qapi/qmp/qdict.h" @@ -467,6 +468,8 @@ typedef struct { bool use_shmem; /* only launch the target process */ bool only_target; + /* Use dirty ring if true; dirty logging otherwise */ + bool use_dirty_ring; char *opts_source; char *opts_target; } MigrateStart; @@ -573,11 +576,13 @@ static int test_migrate_start(QTestState **from, QTestState **to, shmem_opts = g_strdup(""); } - cmd_source = g_strdup_printf("-accel kvm -accel tcg%s%s " + cmd_source = g_strdup_printf("-accel kvm%s -accel tcg%s%s " "-name source,debug-threads=on " "-m %s " "-serial file:%s/src_serial " "%s %s %s %s", + args->use_dirty_ring ? + ",dirty-ring-size=4096" : "", machine_opts ? " -machine " : "", machine_opts ? machine_opts : "", memory_size, tmpfs, @@ -587,12 +592,14 @@ static int test_migrate_start(QTestState **from, QTestState **to, *from = qtest_init(cmd_source); } - cmd_target = g_strdup_printf("-accel kvm -accel tcg%s%s " + cmd_target = g_strdup_printf("-accel kvm%s -accel tcg%s%s " "-name target,debug-threads=on " "-m %s " "-serial file:%s/dest_serial " "-incoming %s " "%s %s %s %s", + args->use_dirty_ring ? + ",dirty-ring-size=4096" : "", machine_opts ? " -machine " : "", machine_opts ? machine_opts : "", memory_size, tmpfs, uri, @@ -785,12 +792,14 @@ static void test_baddest(void) test_migrate_end(from, to, false); } -static void test_precopy_unix(void) +static void test_precopy_unix_common(bool dirty_ring) { g_autofree char *uri = g_strdup_printf("unix:%s/migsocket", tmpfs); MigrateStart *args = migrate_start_new(); QTestState *from, *to; + args->use_dirty_ring = dirty_ring; + if (test_migrate_start(&from, &to, uri, args)) { return; } @@ -825,6 +834,18 @@ static void test_precopy_unix(void) test_migrate_end(from, to, true); } +static void test_precopy_unix(void) +{ + /* Using default dirty logging */ + test_precopy_unix_common(false); +} + +static void test_precopy_unix_dirty_ring(void) +{ + /* Using dirty ring tracking */ + test_precopy_unix_common(true); +} + #if 0 /* Currently upset on aarch64 TCG */ static void test_ignore_shared(void) @@ -1369,6 +1390,25 @@ static void test_multifd_tcp_cancel(void) test_migrate_end(from, to2, true); } +static bool kvm_dirty_ring_supported(void) +{ + int ret, kvm_fd = open("/dev/kvm", O_RDONLY); + + if (kvm_fd < 0) { + return false; + } + + ret = ioctl(kvm_fd, KVM_CHECK_EXTENSION, KVM_CAP_DIRTY_LOG_RING); + close(kvm_fd); + + /* We test with 4096 slots */ + if (ret < 4096) { + return false; + } + + return true; +} + int main(int argc, char **argv) { char template[] = "/tmp/migration-test-XXXXXX"; @@ -1438,6 +1478,11 @@ int main(int argc, char **argv) qtest_add_func("/migration/multifd/tcp/zstd", test_multifd_tcp_zstd); #endif + if (kvm_dirty_ring_supported()) { + qtest_add_func("/migration/dirty_ring", + test_precopy_unix_dirty_ring); + } + ret = g_test_run(); g_assert_cmpint(ret, ==, 0); -- 2.31.1 ^ permalink raw reply related [flat|nested] 7+ messages in thread
* Re: [PATCH 2/2] tests: migration-test: Add dirty ring test 2021-06-10 0:52 ` [PATCH 2/2] tests: migration-test: Add dirty ring test Peter Xu @ 2021-06-10 19:01 ` Dr. David Alan Gilbert 2021-06-10 19:32 ` Peter Xu 0 siblings, 1 reply; 7+ messages in thread From: Dr. David Alan Gilbert @ 2021-06-10 19:01 UTC (permalink / raw) To: Peter Xu Cc: Paolo Bonzini, Leonardo Bras Soares Passos, qemu-devel, Juan Quintela * Peter Xu (peterx@redhat.com) wrote: > Add dirty ring test if kernel supports it. Add the dirty ring parameter on > source should be mostly enough, but let's change the dest too to make them > match always. > > Signed-off-by: Peter Xu <peterx@redhat.com> > --- > tests/qtest/migration-test.c | 51 +++++++++++++++++++++++++++++++++--- > 1 file changed, 48 insertions(+), 3 deletions(-) > > diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c > index d9225f58d4d..cc6e396d1a2 100644 > --- a/tests/qtest/migration-test.c > +++ b/tests/qtest/migration-test.c > @@ -12,6 +12,7 @@ > > #include "qemu/osdep.h" > > +#include <linux/kvm.h> Does that get you the system headers, which may or may not have KVM_CAP_DIRTY_LOG_RING if you're on an old host, or does it get you qemu's linux-headers which definitely does? What happens on a BSD or the like? Dave > #include "libqos/libqtest.h" > #include "qapi/error.h" > #include "qapi/qmp/qdict.h" > @@ -467,6 +468,8 @@ typedef struct { > bool use_shmem; > /* only launch the target process */ > bool only_target; > + /* Use dirty ring if true; dirty logging otherwise */ > + bool use_dirty_ring; > char *opts_source; > char *opts_target; > } MigrateStart; > @@ -573,11 +576,13 @@ static int test_migrate_start(QTestState **from, QTestState **to, > shmem_opts = g_strdup(""); > } > > - cmd_source = g_strdup_printf("-accel kvm -accel tcg%s%s " > + cmd_source = g_strdup_printf("-accel kvm%s -accel tcg%s%s " > "-name source,debug-threads=on " > "-m %s " > "-serial file:%s/src_serial " > "%s %s %s %s", > + args->use_dirty_ring ? > + ",dirty-ring-size=4096" : "", > machine_opts ? " -machine " : "", > machine_opts ? machine_opts : "", > memory_size, tmpfs, > @@ -587,12 +592,14 @@ static int test_migrate_start(QTestState **from, QTestState **to, > *from = qtest_init(cmd_source); > } > > - cmd_target = g_strdup_printf("-accel kvm -accel tcg%s%s " > + cmd_target = g_strdup_printf("-accel kvm%s -accel tcg%s%s " > "-name target,debug-threads=on " > "-m %s " > "-serial file:%s/dest_serial " > "-incoming %s " > "%s %s %s %s", > + args->use_dirty_ring ? > + ",dirty-ring-size=4096" : "", > machine_opts ? " -machine " : "", > machine_opts ? machine_opts : "", > memory_size, tmpfs, uri, > @@ -785,12 +792,14 @@ static void test_baddest(void) > test_migrate_end(from, to, false); > } > > -static void test_precopy_unix(void) > +static void test_precopy_unix_common(bool dirty_ring) > { > g_autofree char *uri = g_strdup_printf("unix:%s/migsocket", tmpfs); > MigrateStart *args = migrate_start_new(); > QTestState *from, *to; > > + args->use_dirty_ring = dirty_ring; > + > if (test_migrate_start(&from, &to, uri, args)) { > return; > } > @@ -825,6 +834,18 @@ static void test_precopy_unix(void) > test_migrate_end(from, to, true); > } > > +static void test_precopy_unix(void) > +{ > + /* Using default dirty logging */ > + test_precopy_unix_common(false); > +} > + > +static void test_precopy_unix_dirty_ring(void) > +{ > + /* Using dirty ring tracking */ > + test_precopy_unix_common(true); > +} > + > #if 0 > /* Currently upset on aarch64 TCG */ > static void test_ignore_shared(void) > @@ -1369,6 +1390,25 @@ static void test_multifd_tcp_cancel(void) > test_migrate_end(from, to2, true); > } > > +static bool kvm_dirty_ring_supported(void) > +{ > + int ret, kvm_fd = open("/dev/kvm", O_RDONLY); > + > + if (kvm_fd < 0) { > + return false; > + } > + > + ret = ioctl(kvm_fd, KVM_CHECK_EXTENSION, KVM_CAP_DIRTY_LOG_RING); > + close(kvm_fd); > + > + /* We test with 4096 slots */ > + if (ret < 4096) { > + return false; > + } > + > + return true; > +} > + > int main(int argc, char **argv) > { > char template[] = "/tmp/migration-test-XXXXXX"; > @@ -1438,6 +1478,11 @@ int main(int argc, char **argv) > qtest_add_func("/migration/multifd/tcp/zstd", test_multifd_tcp_zstd); > #endif > > + if (kvm_dirty_ring_supported()) { > + qtest_add_func("/migration/dirty_ring", > + test_precopy_unix_dirty_ring); > + } > + > ret = g_test_run(); > > g_assert_cmpint(ret, ==, 0); > -- > 2.31.1 > -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/2] tests: migration-test: Add dirty ring test 2021-06-10 19:01 ` Dr. David Alan Gilbert @ 2021-06-10 19:32 ` Peter Xu 2021-06-14 17:55 ` Dr. David Alan Gilbert 0 siblings, 1 reply; 7+ messages in thread From: Peter Xu @ 2021-06-10 19:32 UTC (permalink / raw) To: Dr. David Alan Gilbert Cc: Paolo Bonzini, Leonardo Bras Soares Passos, qemu-devel, Juan Quintela On Thu, Jun 10, 2021 at 08:01:44PM +0100, Dr. David Alan Gilbert wrote: > > +#include <linux/kvm.h> > > Does that get you the system headers, which may or may not have > KVM_CAP_DIRTY_LOG_RING if you're on an old host, or does it get you > qemu's linux-headers which definitely does? I tested it and it's using the linux-headers/ file even if I also got the other /usr/include one. So I think the qemu one just has higher priority in the "-I" paths. Btw, IIUC quotting with <> or "" should be the same here for headers, so I'm thinking maybe I should switch to "" like the rest headers. > > What happens on a BSD or the like? Ah, good point.. How about I squash this into the patch? ---8<--- diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c index cc6e396d1a2..9ef6b471353 100644 --- a/tests/qtest/migration-test.c +++ b/tests/qtest/migration-test.c @@ -12,7 +12,6 @@ #include "qemu/osdep.h" -#include <linux/kvm.h> #include "libqos/libqtest.h" #include "qapi/error.h" #include "qapi/qmp/qdict.h" @@ -28,6 +27,10 @@ #include "migration-helpers.h" #include "tests/migration/migration-test.h" +#if defined(__linux__) +#include "linux/kvm.h" +#endif + /* TODO actually test the results and get rid of this */ #define qtest_qmp_discard_response(...) qobject_unref(qtest_qmp(__VA_ARGS__)) @@ -1392,6 +1395,7 @@ static void test_multifd_tcp_cancel(void) static bool kvm_dirty_ring_supported(void) { +#if defined(__linux__) int ret, kvm_fd = open("/dev/kvm", O_RDONLY); if (kvm_fd < 0) { @@ -1407,6 +1411,9 @@ static bool kvm_dirty_ring_supported(void) } return true; +#else + return false; +#endif } int main(int argc, char **argv) ---8<--- Thanks! -- Peter Xu ^ permalink raw reply related [flat|nested] 7+ messages in thread
* Re: [PATCH 2/2] tests: migration-test: Add dirty ring test 2021-06-10 19:32 ` Peter Xu @ 2021-06-14 17:55 ` Dr. David Alan Gilbert 0 siblings, 0 replies; 7+ messages in thread From: Dr. David Alan Gilbert @ 2021-06-14 17:55 UTC (permalink / raw) To: Peter Xu Cc: Paolo Bonzini, Leonardo Bras Soares Passos, qemu-devel, Juan Quintela * Peter Xu (peterx@redhat.com) wrote: > On Thu, Jun 10, 2021 at 08:01:44PM +0100, Dr. David Alan Gilbert wrote: > > > +#include <linux/kvm.h> > > > > Does that get you the system headers, which may or may not have > > KVM_CAP_DIRTY_LOG_RING if you're on an old host, or does it get you > > qemu's linux-headers which definitely does? > > I tested it and it's using the linux-headers/ file even if I also got the other > /usr/include one. So I think the qemu one just has higher priority in the "-I" > paths. > > Btw, IIUC quotting with <> or "" should be the same here for headers, so I'm > thinking maybe I should switch to "" like the rest headers. Oh in that case, Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com> > > > > > What happens on a BSD or the like? > > Ah, good point.. > > How about I squash this into the patch? I'm not sure; I was only really worried about the build failing, I was assuming that nothing else would have /dev/kvm so your open would fail. Although I guess that might mean something entirely different on some other OS, so perhaps it's worth it. Dave > ---8<--- > diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c > index cc6e396d1a2..9ef6b471353 100644 > --- a/tests/qtest/migration-test.c > +++ b/tests/qtest/migration-test.c > @@ -12,7 +12,6 @@ > > #include "qemu/osdep.h" > > -#include <linux/kvm.h> > #include "libqos/libqtest.h" > #include "qapi/error.h" > #include "qapi/qmp/qdict.h" > @@ -28,6 +27,10 @@ > #include "migration-helpers.h" > #include "tests/migration/migration-test.h" > > +#if defined(__linux__) > +#include "linux/kvm.h" > +#endif > + > /* TODO actually test the results and get rid of this */ > #define qtest_qmp_discard_response(...) qobject_unref(qtest_qmp(__VA_ARGS__)) > > @@ -1392,6 +1395,7 @@ static void test_multifd_tcp_cancel(void) > > static bool kvm_dirty_ring_supported(void) > { > +#if defined(__linux__) > int ret, kvm_fd = open("/dev/kvm", O_RDONLY); > > if (kvm_fd < 0) { > @@ -1407,6 +1411,9 @@ static bool kvm_dirty_ring_supported(void) > } > > return true; > +#else > + return false; > +#endif > } > > int main(int argc, char **argv) > ---8<--- > > Thanks! > > -- > Peter Xu > -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK ^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2021-06-14 17:56 UTC | newest] Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2021-06-10 0:52 [PATCH 0/2] tests: migration-test: Fix agressive test skip, add dirty ring test Peter Xu 2021-06-10 0:52 ` [PATCH 1/2] tests: migration-test: Still run the rest even if uffd missing Peter Xu 2021-06-10 18:56 ` Dr. David Alan Gilbert 2021-06-10 0:52 ` [PATCH 2/2] tests: migration-test: Add dirty ring test Peter Xu 2021-06-10 19:01 ` Dr. David Alan Gilbert 2021-06-10 19:32 ` Peter Xu 2021-06-14 17:55 ` Dr. David Alan Gilbert
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.