pie/restorer: unregister (g)libc rseq before memory restoration

Fresh glibc does rseq registration by default during start_thread().
[ see https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=95e114a0919d844d8fe07839cb6538b7f5ee920e ]

This cause process crashes during memory restore procedure, because
memory which corresponds to the struct rseq will be unmapped and overriden
in __export_restore_task.

Let's perform rseq unregistration just before unmap_old_vmas(). To achieve
that we need to determine (struct rseq) address at first while we are in Glibc
(we do that in prep_libc_rseq_info using Glibc exported symbols).

See also
("nptl: Add public rseq symbols and <sys/rseq.h>")
https://sourceware.org/git?p=glibc.git;a=commit;h=c901c3e764d7c7079f006b4e21e877d5036eb4f5
("nptl: Add <thread_pointer.h> for defining __thread_pointer")
https://sourceware.org/git?p=glibc.git;a=commit;h=8dbeb0561eeb876f557ac9eef5721912ec074ea5

TODO: do the same for musl-libc if it will start to register rseq by default

Signed-off-by: Alexander Mikhalitsyn <alexander.mikhalitsyn@virtuozzo.com>

criu/cr-restore.c

@@ -23,6 +23,7 @@
 #include "common/compiler.h"
 
 #include "linux/mount.h"
+#include "linux/rseq.h"
 
 #include "clone-noasan.h"
 #include "cr_options.h"
@@ -3012,6 +3013,32 @@ static int prep_rseq(struct rst_rseq_param *rseq, ThreadCoreEntry *tc)
 	return 0;
 }
 
+#if defined(__GLIBC__) && defined(RSEQ_SIG)
+static void prep_libc_rseq_info(struct rst_rseq_param *rseq)
+{
+	if (!kdat.has_rseq) {
+		rseq->rseq_abi_pointer = 0;
+		return;
+	}
+
+	rseq->rseq_abi_pointer = encode_pointer(__criu_thread_pointer() + __rseq_offset);
+	rseq->rseq_abi_size = __rseq_size;
+	rseq->signature = RSEQ_SIG;
+}
+#else
+static void prep_libc_rseq_info(struct rst_rseq_param *rseq)
+{
+	/*
+	 * TODO: handle built-in rseq on other libc'ies like musl
+	 * We can do that using get_rseq_conf kernel feature.
+	 *
+	 * For now we just assume that other libc libraries are
+	 * not registering rseq by default.
+	 */
+	rseq->rseq_abi_pointer = 0;
+}
+#endif
+
 static rlim_t decode_rlim(rlim_t ival)
 {
 	return ival == -1 ? RLIM_INFINITY : ival;
@@ -3665,6 +3692,8 @@ static int sigreturn_restore(pid_t pid, struct task_restore_args *task_args, uns
 	strncpy(task_args->comm, core->tc->comm, TASK_COMM_LEN - 1);
 	task_args->comm[TASK_COMM_LEN - 1] = 0;
 
+	prep_libc_rseq_info(&task_args->libc_rseq);
+
 	/*
 	 * Fill up per-thread data.
 	 */

criu/include/linux/rseq.h

@@ -2,6 +2,14 @@
 #ifndef _UAPI_LINUX_RSEQ_H
 #define _UAPI_LINUX_RSEQ_H
 
+#ifdef __has_include
+#if __has_include("sys/rseq.h")
+#include <sys/rseq.h>
+#include "asm/thread_pointer.h"
+#endif
+#endif
+
+#ifndef __GLIBC_HAVE_KERNEL_RSEQ
 /*
  * linux/rseq.h
  *
@@ -49,6 +57,7 @@ struct rseq_cs {
 	__u64 post_commit_offset;
 	__u64 abort_ip;
 } __attribute__((aligned(4 * sizeof(__u64))));
+#endif /* __GLIBC_HAVE_KERNEL_RSEQ */
 
 /*
  * We have to have our own copy of struct rseq definition because

criu/include/restorer.h

@@ -229,6 +229,12 @@ struct task_restore_args {
 	int lsm_type;
 	int child_subreaper;
 	bool has_clone3_set_tid;
+
+	/*
+	 * info about rseq from libc used to
+	 * unregister it before memory restoration procedure
+	 */
+	struct rst_rseq_param libc_rseq;
 } __aligned(64);
 
 /*

criu/pie/restorer.c

@@ -1122,6 +1122,15 @@ void __export_unmap(void)
 	sys_munmap(bootstrap_start, bootstrap_len - vdso_rt_size);
 }
 
+static void unregister_libc_rseq(struct rst_rseq_param *rseq)
+{
+	if (!rseq->rseq_abi_pointer)
+		return;
+
+	/* can't fail if rseq is registered */
+	sys_rseq(decode_pointer(rseq->rseq_abi_pointer), rseq->rseq_abi_size, 1, rseq->signature);
+}
+
 /*
  * This function unmaps all VMAs, which don't belong to
  * the restored process or the restorer.
@@ -1461,6 +1470,15 @@ long __export_restore_task(struct task_restore_args *args)
 			goto core_restore_end;
 	}
 
+	/*
+	 * We may have rseq registered already if CRIU compiled against
+	 * a fresh Glibc with rseq support. Anyway, we need to unregister it
+	 * before doing unmap_old_vmas or we will get SIGSEGV from the kernel,
+	 * for instance once the kernel will want to update (struct rseq).cpu_id field:
+	 * https://github.com/torvalds/linux/blob/ce522ba9ef7e/kernel/rseq.c#L89
+	 */
+	unregister_libc_rseq(&args->libc_rseq);
+
 	if (unmap_old_vmas((void *)args->premmapped_addr, args->premmapped_len, bootstrap_start, bootstrap_len,
 			   args->task_size))
 		goto core_restore_end;