diff mbox series

[v12,12/13] KVM: selftests: Let dirty_log_test async for dirty ring test

Message ID 20200930215449.48343-1-peterx@redhat.com (mailing list archive)
State New, archived
Headers show
Series KVM: Dirty ring interface | expand

Commit Message

Peter Xu Sept. 30, 2020, 9:54 p.m. UTC 
Previously the dirty ring test was working in synchronous way, because
only with a vmexit (with that it was the ring full event) we'll know
the hardware dirty bits will be flushed to the dirty ring.

With this patch we first introduced the vcpu kick mechanism by using
SIGUSR1, meanwhile we can have a guarantee of vmexit and also the
flushing of hardware dirty bits.  With all these, we can keep the vcpu
dirty work asynchronous of the whole collection procedure now.  Still,
we need to be very careful that we can only do it async if the vcpu is
not reaching soft limit (no KVM_EXIT_DIRTY_RING_FULL).  Otherwise we
must collect the dirty bits before continuing the vcpu.

Further increase the dirty ring size to current maximum to make sure
we torture more on the no-ring-full case, which should be the major
scenario when the hypervisors like QEMU would like to use this feature.

Reviewed-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
---
 tools/testing/selftests/kvm/dirty_log_test.c  | 126 +++++++++++++-----
 .../testing/selftests/kvm/include/kvm_util.h  |   1 +
 tools/testing/selftests/kvm/lib/kvm_util.c    |   9 ++
 3 files changed, 106 insertions(+), 30 deletions(-)

Comments

Peter Xu Oct. 7, 2020, 8:45 p.m. UTC | #1 
On Sun, Oct 04, 2020 at 09:26:31PM +0800, kernel test robot wrote:
> # fetch 0x1fff3 page 13105==== Test Assertion Failure ====
> #   dirty_log_test.c:526: test_bit_le(page, bmap)
> #   pid=10583 tid=10583 - Invalid argument
> #      1	0x0000000000403294: vm_dirty_log_verify at dirty_log_test.c:523
> #      2	 (inlined by) run_test at dirty_log_test.c:743
> #      3	0x00000000004026f4: main at dirty_log_test.c:905 (discriminator 3)
> #      4	0x00007f2d11a2809a: ?? ??:0
> #      5	0x0000000000402809: _start at ??:?
> #   Page 131151 should have its dirty bit set in this iteration but it is missing

I think I reproduced this error even on my latest tree.  I don't think it's a
kvm bug, instead it's a bug in the test case.  Not too surprising considering
that I spent probably more time on the test case than the kernel changes.

KVM dirty ring introduced a chance for spurious dirty bit (safe, but it just
made the test even more tricky) which does not exist for dirty logging.  It
happens when the ring gets full and the last PFN written could be spurious,
simply because when kvm noticed that the ring is full, it won't continue the
guest (whose next instruction is to do the real RAM update) but instead it'll
quit to userspace with that page PFN actually set but with old data.

The current test case is very strict on checking stuff. It tries to skip this
magic bit so far by recording it into host_bmap_track so we don't check the
data in this round, but we expect this bit to be set in the next run (because
it's the next guest instruction that gonna happen soon).  However I just
noticed that when accidentally the vcpu thread didn't really continue (e.g.,
when the collection thread is too slow, due to when we enabled all debug
prints) then we won't find that special bit in the next iteration of dirty
bitmap simply because the vcpu has already stopped and never go into the next
round.

Anyway, I'm posting the fix here for reference which have worked for me.  Below
is the thing I plan to squash into the test patch (probably not the current
one, but previous one):

-------------8<--------------------
diff --git a/tools/testing/selftests/kvm/dirty_log_test.c b/tools/testing/selftests/kvm/dirty_log_test.c
index 80c42c87265e..aecd0f99f13c 100644
--- a/tools/testing/selftests/kvm/dirty_log_test.c
+++ b/tools/testing/selftests/kvm/dirty_log_test.c
@@ -146,6 +146,20 @@ static sem_t dirty_ring_vcpu_cont;
  * dirty_ring_vcpu_stop and before vcpu continues to run.
  */
 static bool dirty_ring_vcpu_ring_full;
+/*
+ * This is only used for verifying the dirty pages.  Dirty ring has a very
+ * tricky case when the ring just got full, kvm will do userspace exit due to
+ * ring full.  When that happens, the very last PFN is very tricky in that it's
+ * set but actually the data is not changed (the guest WRITE is not really
+ * applied yet), because when we find that the dirty ring is full, we refused
+ * to continue the vcpu, hence we got the dirty gfn recorded without the new
+ * data.  For this specific case, it's safe to skip checking this pfn for this
+ * bit, because it's a redundant bit, and when the write happens later the bit
+ * will be set again.  We use this variable to always keep track of the latest
+ * dirty gfn we've collected, so that if a mismatch of data found later in the
+ * verifying process, we let it pass.
+ */
+static uint64_t dirty_ring_last_page;

 enum log_mode_t {
        /* Only use KVM_GET_DIRTY_LOG for logging */
@@ -281,7 +295,8 @@ static uint32_t dirty_ring_collect_one(struct kvm_dirty_gfn *dirty_gfns,
                TEST_ASSERT(cur->offset < num_pages, "Offset overflow: "
                            "0x%llx >= 0x%x", cur->offset, num_pages);
                pr_info("fetch 0x%x page %llu\n", *fetch_index, cur->offset);
-               set_bit(cur->offset, bitmap);
+               set_bit_le(cur->offset, bitmap);
+               dirty_ring_last_page = cur->offset;
                dirty_gfn_set_collected(cur);
                (*fetch_index)++;
                count++;
@@ -572,16 +587,12 @@ static void vm_dirty_log_verify(enum vm_guest_mode mode, unsigned long *bmap)
                                         *    (here if we read value on page
                                         *     "2R-2" is 1, while iter=3!!!)
                                         */
-                                       matched = true;
-                               } else {
+                                       continue;
+                               } else if (page == dirty_ring_last_page) {
                                        /*
-                                        * This is also special for dirty ring
-                                        * when this page is exactly the last
-                                        * page touched before vcpu ring full.
-                                        * If it happens, we should expect the
-                                        * value to change in the next round.
+                                        * Please refer to comments in
+                                        * dirty_ring_last_page.
                                         */
-                                       set_bit_le(page, host_bmap_track);
                                        continue;
                                }
                        }
-------------8<--------------------

Majorly what I did is:

  - A new variable 'dirty_ring_last_page' is introduced to the test to record
    the last dirtied GFN.  When verifying the dirty bits, if we found a
    spurious dirty bit on this specific GFN, we skip.  This should be more
    solid than the previous trick to pass it onto host_bmap_track.

  - Since at it, I touched up two other places:

    - Replaced another set_bit() that I just found, which would otherwise break
      ppc I believe (but I must also confess I don't know when ppc will support
      dirty ring)

    - Change one "matched = true;" into a "continue;".  It should do the same
      thing, but "continue" is more obvious.

I'll resend the series.  I won't touch anything else, but just to (hopefully)
trigger the testbot again on the whole series.
diff mbox series

Patch

diff --git a/tools/testing/selftests/kvm/dirty_log_test.c b/tools/testing/selftests/kvm/dirty_log_test.c
index 531431cff4fc..4b404dfdc2f9 100644
--- a/tools/testing/selftests/kvm/dirty_log_test.c
+++ b/tools/testing/selftests/kvm/dirty_log_test.c
@@ -13,6 +13,9 @@ 
 #include <time.h>
 #include <pthread.h>
 #include <semaphore.h>
+#include <sys/types.h>
+#include <signal.h>
+#include <errno.h>
 #include <linux/bitmap.h>
 #include <linux/bitops.h>
 #include <asm/barrier.h>
@@ -59,7 +62,9 @@ 
 # define test_and_clear_bit_le	test_and_clear_bit
 #endif
 
-#define TEST_DIRTY_RING_COUNT		1024
+#define TEST_DIRTY_RING_COUNT		65536
+
+#define SIG_IPI SIGUSR1
 
 /*
  * Guest/Host shared variables. Ensure addr_gva2hva() and/or
@@ -135,6 +140,12 @@  static uint64_t host_track_next_count;
 /* Whether dirty ring reset is requested, or finished */
 static sem_t dirty_ring_vcpu_stop;
 static sem_t dirty_ring_vcpu_cont;
+/*
+ * This is updated by the vcpu thread to tell the host whether it's a
+ * ring-full event.  It should only be read until a sem_wait() of
+ * dirty_ring_vcpu_stop and before vcpu continues to run.
+ */
+static bool dirty_ring_vcpu_ring_full;
 
 enum log_mode_t {
 	/* Only use KVM_GET_DIRTY_LOG for logging */
@@ -156,6 +167,33 @@  enum log_mode_t {
 static enum log_mode_t host_log_mode_option = LOG_MODE_ALL;
 /* Logging mode for current run */
 static enum log_mode_t host_log_mode;
+static pthread_t vcpu_thread;
+
+/* Only way to pass this to the signal handler */
+static struct kvm_vm *current_vm;
+
+static void vcpu_sig_handler(int sig)
+{
+	TEST_ASSERT(sig == SIG_IPI, "unknown signal: %d", sig);
+}
+
+static void vcpu_kick(void)
+{
+	pthread_kill(vcpu_thread, SIG_IPI);
+}
+
+/*
+ * In our test we do signal tricks, let's use a better version of
+ * sem_wait to avoid signal interrupts
+ */
+static void sem_wait_until(sem_t *sem)
+{
+	int ret;
+
+	do
+		ret = sem_wait(sem);
+	while (ret == -1 && errno == EINTR);
+}
 
 static bool clear_log_supported(void)
 {
@@ -189,10 +227,13 @@  static void clear_log_collect_dirty_pages(struct kvm_vm *vm, int slot,
 	kvm_vm_clear_dirty_log(vm, slot, bitmap, 0, num_pages);
 }
 
-static void default_after_vcpu_run(struct kvm_vm *vm)
+static void default_after_vcpu_run(struct kvm_vm *vm, int ret, int err)
 {
 	struct kvm_run *run = vcpu_state(vm, VCPU_ID);
 
+	TEST_ASSERT(ret == 0 || (ret == -1 && err == EINTR),
+		    "vcpu run failed: errno=%d", err);
+
 	TEST_ASSERT(get_ucall(vm, VCPU_ID, NULL) == UCALL_SYNC,
 		    "Invalid guest sync status: exit_reason=%s\n",
 		    exit_reason_str(run->exit_reason));
@@ -248,27 +289,37 @@  static uint32_t dirty_ring_collect_one(struct kvm_dirty_gfn *dirty_gfns,
 	return count;
 }
 
+static void dirty_ring_wait_vcpu(void)
+{
+	/* This makes sure that hardware PML cache flushed */
+	vcpu_kick();
+	sem_wait_until(&dirty_ring_vcpu_stop);
+}
+
+static void dirty_ring_continue_vcpu(void)
+{
+	pr_info("Notifying vcpu to continue\n");
+	sem_post(&dirty_ring_vcpu_cont);
+}
+
 static void dirty_ring_collect_dirty_pages(struct kvm_vm *vm, int slot,
 					   void *bitmap, uint32_t num_pages)
 {
 	/* We only have one vcpu */
 	static uint32_t fetch_index = 0;
 	uint32_t count = 0, cleared;
+	bool continued_vcpu = false;
 
-	/*
-	 * Before fetching the dirty pages, we need a vmexit of the
-	 * worker vcpu to make sure the hardware dirty buffers were
-	 * flushed.  This is not needed for dirty-log/clear-log tests
-	 * because get dirty log will natually do so.
-	 *
-	 * For now we do it in the simple way - we simply wait until
-	 * the vcpu uses up the soft dirty ring, then it'll always
-	 * do a vmexit to make sure that PML buffers will be flushed.
-	 * In real hypervisors, we probably need a vcpu kick or to
-	 * stop the vcpus (before the final sync) to make sure we'll
-	 * get all the existing dirty PFNs even cached in hardware.
-	 */
-	sem_wait(&dirty_ring_vcpu_stop);
+	dirty_ring_wait_vcpu();
+
+	if (!dirty_ring_vcpu_ring_full) {
+		/*
+		 * This is not a ring-full event, it's safe to allow
+		 * vcpu to continue
+		 */
+		dirty_ring_continue_vcpu();
+		continued_vcpu = true;
+	}
 
 	/* Only have one vcpu */
 	count = dirty_ring_collect_one(vcpu_map_dirty_ring(vm, VCPU_ID),
@@ -280,13 +331,16 @@  static void dirty_ring_collect_dirty_pages(struct kvm_vm *vm, int slot,
 	TEST_ASSERT(cleared == count, "Reset dirty pages (%u) mismatch "
 		    "with collected (%u)", cleared, count);
 
-	pr_info("Notifying vcpu to continue\n");
-	sem_post(&dirty_ring_vcpu_cont);
+	if (!continued_vcpu) {
+		TEST_ASSERT(dirty_ring_vcpu_ring_full,
+			    "Didn't continue vcpu even without ring full");
+		dirty_ring_continue_vcpu();
+	}
 
 	pr_info("Iteration %ld collected %u pages\n", iteration, count);
 }
 
-static void dirty_ring_after_vcpu_run(struct kvm_vm *vm)
+static void dirty_ring_after_vcpu_run(struct kvm_vm *vm, int ret, int err)
 {
 	struct kvm_run *run = vcpu_state(vm, VCPU_ID);
 
@@ -294,10 +348,16 @@  static void dirty_ring_after_vcpu_run(struct kvm_vm *vm)
 	if (get_ucall(vm, VCPU_ID, NULL) == UCALL_SYNC) {
 		/* We should allow this to continue */
 		;
-	} else if (run->exit_reason == KVM_EXIT_DIRTY_RING_FULL) {
+	} else if (run->exit_reason == KVM_EXIT_DIRTY_RING_FULL ||
+		   (ret == -1 && err == EINTR)) {
+		/* Update the flag first before pause */
+		WRITE_ONCE(dirty_ring_vcpu_ring_full,
+			   run->exit_reason == KVM_EXIT_DIRTY_RING_FULL);
 		sem_post(&dirty_ring_vcpu_stop);
-		pr_info("vcpu stops because dirty ring full...\n");
-		sem_wait(&dirty_ring_vcpu_cont);
+		pr_info("vcpu stops because %s...\n",
+			dirty_ring_vcpu_ring_full ?
+			"dirty ring is full" : "vcpu is kicked out");
+		sem_wait_until(&dirty_ring_vcpu_cont);
 		pr_info("vcpu continues now.\n");
 	} else {
 		TEST_ASSERT(false, "Invalid guest sync status: "
@@ -322,7 +382,7 @@  struct log_mode {
 	void (*collect_dirty_pages) (struct kvm_vm *vm, int slot,
 				     void *bitmap, uint32_t num_pages);
 	/* Hook to call when after each vcpu run */
-	void (*after_vcpu_run)(struct kvm_vm *vm);
+	void (*after_vcpu_run)(struct kvm_vm *vm, int ret, int err);
 	void (*before_vcpu_join) (void);
 } log_modes[LOG_MODE_NUM] = {
 	{
@@ -394,12 +454,12 @@  static void log_mode_collect_dirty_pages(struct kvm_vm *vm, int slot,
 	mode->collect_dirty_pages(vm, slot, bitmap, num_pages);
 }
 
-static void log_mode_after_vcpu_run(struct kvm_vm *vm)
+static void log_mode_after_vcpu_run(struct kvm_vm *vm, int ret, int err)
 {
 	struct log_mode *mode = &log_modes[host_log_mode];
 
 	if (mode->after_vcpu_run)
-		mode->after_vcpu_run(vm);
+		mode->after_vcpu_run(vm, ret, err);
 }
 
 static void log_mode_before_vcpu_join(void)
@@ -420,20 +480,27 @@  static void generate_random_array(uint64_t *guest_array, uint64_t size)
 
 static void *vcpu_worker(void *data)
 {
-	int ret;
+	int ret, vcpu_fd;
 	struct kvm_vm *vm = data;
 	uint64_t *guest_array;
 	uint64_t pages_count = 0;
+	struct sigaction sigact;
+
+	current_vm = vm;
+	vcpu_fd = vcpu_get_fd(vm, VCPU_ID);
+	memset(&sigact, 0, sizeof(sigact));
+	sigact.sa_handler = vcpu_sig_handler;
+	sigaction(SIG_IPI, &sigact, NULL);
 
 	guest_array = addr_gva2hva(vm, (vm_vaddr_t)random_array);
 
 	while (!READ_ONCE(host_quit)) {
+		/* Clear any existing kick signals */
 		generate_random_array(guest_array, TEST_PAGES_PER_LOOP);
 		pages_count += TEST_PAGES_PER_LOOP;
 		/* Let the guest dirty the random pages */
-		ret = _vcpu_run(vm, VCPU_ID);
-		TEST_ASSERT(ret == 0, "vcpu_run failed: %d\n", ret);
-		log_mode_after_vcpu_run(vm);
+		ret = ioctl(vcpu_fd, KVM_RUN, NULL);
+		log_mode_after_vcpu_run(vm, ret, errno);
 	}
 
 	pr_info("Dirtied %"PRIu64" pages\n", pages_count);
@@ -583,7 +650,6 @@  static struct kvm_vm *create_vm(enum vm_guest_mode mode, uint32_t vcpuid,
 static void run_test(enum vm_guest_mode mode, unsigned long iterations,
 		     unsigned long interval, uint64_t phys_offset)
 {
-	pthread_t vcpu_thread;
 	struct kvm_vm *vm;
 	unsigned long *bmap;
 
diff --git a/tools/testing/selftests/kvm/include/kvm_util.h b/tools/testing/selftests/kvm/include/kvm_util.h
index dce07d354a70..f3b5da383bb5 100644
--- a/tools/testing/selftests/kvm/include/kvm_util.h
+++ b/tools/testing/selftests/kvm/include/kvm_util.h
@@ -146,6 +146,7 @@  vm_paddr_t addr_gva2gpa(struct kvm_vm *vm, vm_vaddr_t gva);
 struct kvm_run *vcpu_state(struct kvm_vm *vm, uint32_t vcpuid);
 void vcpu_run(struct kvm_vm *vm, uint32_t vcpuid);
 int _vcpu_run(struct kvm_vm *vm, uint32_t vcpuid);
+int vcpu_get_fd(struct kvm_vm *vm, uint32_t vcpuid);
 void vcpu_run_complete_io(struct kvm_vm *vm, uint32_t vcpuid);
 void vcpu_set_guest_debug(struct kvm_vm *vm, uint32_t vcpuid,
 			  struct kvm_guest_debug *debug);
diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index 831f986674ca..4625e193074e 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -1220,6 +1220,15 @@  int _vcpu_run(struct kvm_vm *vm, uint32_t vcpuid)
 	return rc;
 }
 
+int vcpu_get_fd(struct kvm_vm *vm, uint32_t vcpuid)
+{
+	struct vcpu *vcpu = vcpu_find(vm, vcpuid);
+
+	TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
+
+	return vcpu->fd;
+}
+
 void vcpu_run_complete_io(struct kvm_vm *vm, uint32_t vcpuid)
 {
 	struct vcpu *vcpu = vcpu_find(vm, vcpuid);