[PATCH 3/3] kvmtool/x86: implement guest_pre_init

[Oleg Nesterov <oleg@redhat.com>]

Add the tiny x86/init.S which just mounts /host and execs
/virt/init.

NOTE: of course, the usage of CONFIG_GUEST_PRE_INIT is ugly, we
need to cleanup this code. But I'd prefer to do this on top of
this minimal/simple change. And I think this needs cleanups in
any case, for example I think lkvm shouldn't abuse the "init="
kernel parameter at all.

TODO: x86/init.S doesn't handle i386, should be simple to add.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 Makefile        |  1 +
 builtin-run.c   |  4 ++++
 builtin-setup.c | 14 +++++++++++++-
 guest/init.c    |  2 ++
 x86/init.S      | 38 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 58 insertions(+), 1 deletion(-)
 create mode 100644 x86/init.S

diff --git a/Makefile b/Makefile
index 0f8e003..f8f7cc4 100644
--- a/Makefile
+++ b/Makefile
@@ -110,6 +110,7 @@ endif
 ifeq ($(ARCH),x86_64)
 	ARCH         := x86
 	DEFINES      += -DCONFIG_X86_64
+	ARCH_PRE_INIT = x86/init.S
 endif
 
 ### Arch-specific stuff
diff --git a/builtin-run.c b/builtin-run.c
index e0c8732..3da00da 100644
--- a/builtin-run.c
+++ b/builtin-run.c
@@ -600,7 +600,11 @@ static struct kvm *kvm_cmd_run_init(int argc, const char **argv)
 		if (kvm->cfg.custom_rootfs) {
 			kvm_run_set_sandbox(kvm);
 
+#ifdef CONFIG_GUEST_PRE_INIT
+			strcat(real_cmdline, " init=/virt/pre_init");
+#else
 			strcat(real_cmdline, " init=/virt/init");
+#endif
 
 			if (!kvm->cfg.no_dhcp)
 				strcat(real_cmdline, "  ip=dhcp");
diff --git a/builtin-setup.c b/builtin-setup.c
index bb7c71c..1e5b1e4 100644
--- a/builtin-setup.c
+++ b/builtin-setup.c
@@ -144,12 +144,24 @@ static int extract_file(const char *guestfs_name, const char *filename,
 
 extern char _binary_guest_init_start;
 extern char _binary_guest_init_size;
+extern char _binary_guest_pre_init_start;
+extern char _binary_guest_pre_init_size;
 
 int kvm_setup_guest_init(const char *guestfs_name)
 {
-	return extract_file(guestfs_name, "virt/init",
+	int err;
+
+#ifdef CONFIG_GUEST_PRE_INIT
+	err = extract_file(guestfs_name, "virt/pre_init",
+				&_binary_guest_pre_init_start,
+				&_binary_guest_pre_init_size);
+	if (err)
+		return err;
+#endif
+	err = extract_file(guestfs_name, "virt/init",
 				&_binary_guest_init_start,
 				&_binary_guest_init_size);
+	return err;
 }
 #else
 int kvm_setup_guest_init(const char *guestfs_name)
diff --git a/guest/init.c b/guest/init.c
index 7277a07..46e3fa4 100644
--- a/guest/init.c
+++ b/guest/init.c
@@ -30,7 +30,9 @@ static int run_process_sandbox(char *filename)
 
 static void do_mounts(void)
 {
+#ifndef CONFIG_GUEST_PRE_INIT
 	mount("hostfs", "/host", "9p", MS_RDONLY, "trans=virtio,version=9p2000.L");
+#endif
 	mount("", "/sys", "sysfs", 0, NULL);
 	mount("proc", "/proc", "proc", 0, NULL);
 	mount("devtmpfs", "/dev", "devtmpfs", 0, NULL);
diff --git a/x86/init.S b/x86/init.S
new file mode 100644
index 0000000..488a93f
--- /dev/null
+++ b/x86/init.S
@@ -0,0 +1,38 @@
+.data
+
+.m_dev:
+.string "hostfs"
+.m_dir:
+.string "/host"
+.m_typ:
+.string "9p"
+.m_opt:
+.string "trans=virtio,version=9p2000.L"
+
+.e_nam:
+.string "/virt/init"
+
+.text
+.globl _start
+_start:
+
+	mov $165, %rax		# __NR_mount
+	mov $.m_dev, %rdi
+	mov $.m_dir, %rsi
+	mov $.m_typ, %rdx
+	mov $1, %r10		# MS_RDONLY
+	mov $.m_opt, %r8
+	syscall
+
+	mov $59, %rax		# __NR_execve
+	mov $.e_nam, %rdi
+	lea 8(%rsp), %rsi	# argv[]
+	mov %rdi, (%rsi)	# change argv[0]
+	pop %rcx		# argc
+	inc %rcx
+	lea (%rsi,%rcx,8), %rdx # envp[]
+	syscall
+
+	mov $60, %rax		# __NR_exit
+	mov $1, %rdi
+	syscall			# panic
-- 
2.4.3