Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Gavin Shan <gshan@redhat.com>
To:     kvmarm@lists.cs.columbia.edu
Cc:     kvm@vger.kernel.org, linux-kselftest@vger.kernel.org,
        linux-kernel@vger.kernel.org, pbonzini@redhat.com, maz@kernel.org,
        oliver.upton@linux.dev, andrew.jones@linux.dev, seanjc@google.com,
        mathieu.desnoyers@efficios.com, fweimer@redhat.com,
        yihyu@redhat.com, shan.gavin@gmail.com
Subject: [PATCH 1/2] KVM: selftests: Make rseq compatible with glibc-2.35
Date:   Tue,  9 Aug 2022 14:06:26 +0800
Message-Id: <20220809060627.115847-2-gshan@redhat.com>
In-Reply-To: <20220809060627.115847-1-gshan@redhat.com>
References: <20220809060627.115847-1-gshan@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

The rseq information is registered by TLS, starting from glibc-2.35.
In this case, the test always fails due to syscall(__NR_rseq). For
example, on RHEL9.1 where upstream glibc-2.35 features are enabled
on downstream glibc-2.34, the test fails like below.

  # ./rseq_test
  ==== Test Assertion Failure ====
    rseq_test.c:60: !r
    pid=112043 tid=112043 errno=22 - Invalid argument
       1	0x0000000000401973: main at rseq_test.c:226
       2	0x0000ffff84b6c79b: ?? ??:0
       3	0x0000ffff84b6c86b: ?? ??:0
       4	0x0000000000401b6f: _start at ??:?
    rseq failed, errno = 22 (Invalid argument)
  # rpm -aq | grep glibc-2
  glibc-2.34-39.el9.aarch64

Fix the issue by using the registered rseq information from TLS
if it exists. Otherwise, we're going to register our own rseq
information as before.

Reported-by: Yihuang Yu <yihyu@redhat.com>
Suggested-by: Florian Weimer <fweimer@redhat.com>
Suggested-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Gavin Shan <gshan@redhat.com>
---
 tools/testing/selftests/kvm/rseq_test.c | 30 +++++++++++++++++++++++--
 1 file changed, 28 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/kvm/rseq_test.c b/tools/testing/selftests/kvm/rseq_test.c
index a54d4d05a058..acb1bf1f06b3 100644
--- a/tools/testing/selftests/kvm/rseq_test.c
+++ b/tools/testing/selftests/kvm/rseq_test.c
@@ -9,6 +9,7 @@
 #include <string.h>
 #include <signal.h>
 #include <syscall.h>
+#include <dlfcn.h>
 #include <sys/ioctl.h>
 #include <sys/sysinfo.h>
 #include <asm/barrier.h>
@@ -36,6 +37,8 @@ static __thread volatile struct rseq __rseq = {
  */
 #define NR_TASK_MIGRATIONS 100000
 
+static bool __rseq_ownership;
+static volatile struct rseq *__rseq_info;
 static pthread_t migration_thread;
 static cpu_set_t possible_mask;
 static int min_cpu, max_cpu;
@@ -49,11 +52,33 @@ static void guest_code(void)
 		GUEST_SYNC(0);
 }
 
+static void sys_rseq_ownership(void)
+{
+	long *offset;
+	unsigned int *size, *flags;
+
+	offset = dlsym(RTLD_NEXT, "__rseq_offset");
+	size = dlsym(RTLD_NEXT, "__rseq_size");
+	flags = dlsym(RTLD_NEXT, "__rseq_flags");
+
+	if (offset && size && *size && flags) {
+		__rseq_ownership = false;
+		__rseq_info = (struct rseq *)((uintptr_t)__builtin_thread_pointer() +
+					      *offset);
+	} else {
+		__rseq_ownership = true;
+		__rseq_info = &__rseq;
+	}
+}
+
 static void sys_rseq(int flags)
 {
 	int r;
 
-	r = syscall(__NR_rseq, &__rseq, sizeof(__rseq), flags, RSEQ_SIG);
+	if (!__rseq_ownership)
+		return;
+
+	r = syscall(__NR_rseq, __rseq_info, sizeof(*__rseq_info), flags, RSEQ_SIG);
 	TEST_ASSERT(!r, "rseq failed, errno = %d (%s)", errno, strerror(errno));
 }
 
@@ -218,6 +243,7 @@ int main(int argc, char *argv[])
 
 	calc_min_max_cpu();
 
+	sys_rseq_ownership();
 	sys_rseq(0);
 
 	/*
@@ -256,7 +282,7 @@ int main(int argc, char *argv[])
 			 */
 			smp_rmb();
 			cpu = sched_getcpu();
-			rseq_cpu = READ_ONCE(__rseq.cpu_id);
+			rseq_cpu = READ_ONCE(__rseq_info->cpu_id);
 			smp_rmb();
 		} while (snapshot != atomic_read(&seq_cnt));
 
-- 
2.23.0