[PATCH 0/3] kvmtool: tiny init fox x86_64

[PATCH 0/3] kvmtool: tiny init fox x86_64

[Oleg Nesterov]

Hello,

Yesterday I discovered kvmtool and it looks really cool! Thanks!
Looks like, it does exactly what I need.

But the static /virt/init doesn't look good. This series lessens
the size of lkvm binary from 1045952 to 196200 bytes on x86_64,
but this is minor. I want to write my /virt/init in shell or perl
and this change can help.

I don't really know who should be cc'ed, I've picked some names
from git-log guest/init.c.

Oleg.

 .gitignore      |  1 +
 Makefile        | 26 +++++++++++++++++++++-----
 builtin-run.c   |  4 ++++
 builtin-setup.c | 36 ++++++++++++++++++++++++++----------
 guest/init.c    |  2 ++
 x86/init.S      | 38 ++++++++++++++++++++++++++++++++++++++
 6 files changed, 92 insertions(+), 15 deletions(-)

[PATCH 1/3] kvmtool/setup: Introduce extract_file() helper

[Oleg Nesterov]

Turn kvm_setup_guest_init(guestfs_name) into a more generic helper,
extract_file(guestfs_name, filename, data, size) and reimplement
kvm_setup_guest_init() as a trivial wrapper.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 builtin-setup.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)

diff --git a/builtin-setup.c b/builtin-setup.c
index 40fef15..bb7c71c 100644
--- a/builtin-setup.c
+++ b/builtin-setup.c
@@ -122,30 +122,34 @@ static const char *guestfs_symlinks[] = {
 };
 
 #ifdef CONFIG_GUEST_INIT
-extern char _binary_guest_init_start;
-extern char _binary_guest_init_size;
-
-int kvm_setup_guest_init(const char *guestfs_name)
+static int extract_file(const char *guestfs_name, const char *filename,
+			const void *data, const void *_size)
 {
 	char path[PATH_MAX];
-	size_t size;
 	int fd, ret;
-	char *data;
 
-	size = (size_t)&_binary_guest_init_size;
-	data = (char *)&_binary_guest_init_start;
-	snprintf(path, PATH_MAX, "%s%s/virt/init", kvm__get_dir(), guestfs_name);
+	snprintf(path, PATH_MAX, "%s%s/%s", kvm__get_dir(),
+				guestfs_name, filename);
 	remove(path);
 	fd = open(path, O_CREAT | O_WRONLY, 0755);
 	if (fd < 0)
 		die("Fail to setup %s", path);
-	ret = xwrite(fd, data, size);
+	ret = xwrite(fd, data, (size_t)_size);
 	if (ret < 0)
 		die("Fail to setup %s", path);
 	close(fd);
 
 	return 0;
+}
 
+extern char _binary_guest_init_start;
+extern char _binary_guest_init_size;
+
+int kvm_setup_guest_init(const char *guestfs_name)
+{
+	return extract_file(guestfs_name, "virt/init",
+				&_binary_guest_init_start,
+				&_binary_guest_init_size);
 }
 #else
 int kvm_setup_guest_init(const char *guestfs_name)
-- 
2.4.3

[PATCH 2/3] kvmtool/build: introduce GUEST_PRE_INIT target

[Oleg Nesterov]

This comes as a separate patch because I do not really understand
/usr/bin/make, probably it should be updated.

Change the main Makefile so that if an arch defines ARCH_PRE_INIT
then we

    - build $GUEST_INIT without "-static"

    - add -DCONFIG_GUEST_PRE_INIT to $CFLAGS

    - build $ARCH_PRE_INIT as guest/guest_pre_init.o and embed it
      into lkvm the same as we do with guest/guest_init.o

This also means that ARCH_PRE_INIT case doesn't depend on glibc-static,
we can relax the SOURCE_STATIC check later.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 .gitignore |  1 +
 Makefile   | 25 ++++++++++++++++++++-----
 2 files changed, 21 insertions(+), 5 deletions(-)

diff --git a/.gitignore b/.gitignore
index f10d3c5..697a63f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -9,6 +9,7 @@ include/common-cmds.h
 tests/boot/boot_test.iso
 tests/boot/rootfs/
 guest/init
+guest/pre_init
 guest/init_stage2
 KVMTOOLS-VERSION-FILE
 /x86/bios/bios.bin
diff --git a/Makefile b/Makefile
index f1701aa..0f8e003 100644
--- a/Makefile
+++ b/Makefile
@@ -281,6 +281,14 @@ ifeq ($(call try-build,$(SOURCE_STATIC),,-static),y)
 	CFLAGS		+= -DCONFIG_GUEST_INIT
 	GUEST_INIT	:= guest/init
 	GUEST_OBJS	= guest/guest_init.o
+	ifeq ($(ARCH_PRE_INIT),)
+		GUEST_INIT_FLAGS	+= -static
+	else
+		CFLAGS			+= -DCONFIG_GUEST_PRE_INIT
+		GUEST_INIT_FLAGS	+= -DCONFIG_GUEST_PRE_INIT
+		GUEST_PRE_INIT		:= guest/pre_init
+		GUEST_OBJS		+= guest/guest_pre_init.o
+	endif
 else
 	$(warning No static libc found. Skipping guest init)
 	NOTFOUND        += static-libc
@@ -346,7 +354,7 @@ ifneq ($(WERROR),0)
 	CFLAGS += -Werror
 endif
 
-all: $(PROGRAM) $(PROGRAM_ALIAS) $(GUEST_INIT)
+all: $(PROGRAM) $(PROGRAM_ALIAS) $(GUEST_INIT) $(GUEST_PRE_INIT)
 
 # CFLAGS used when building objects
 # This is intentionally not assigned using :=
@@ -360,11 +368,11 @@ c_flags	= -Wp,-MD,$(depfile) $(CFLAGS)
 #
 STATIC_OBJS = $(patsubst %.o,%.static.o,$(OBJS) $(OBJS_STATOPT))
 
-$(PROGRAM)-static:  $(STATIC_OBJS) $(OTHEROBJS) $(GUEST_INIT)
+$(PROGRAM)-static:  $(STATIC_OBJS) $(OTHEROBJS) $(GUEST_INIT) $(GUEST_PRE_INIT)
 	$(E) "  LINK    " $@
 	$(Q) $(CC) -static $(CFLAGS) $(STATIC_OBJS) $(OTHEROBJS) $(GUEST_OBJS) $(LIBS) $(LIBS_STATOPT) -o $@
 
-$(PROGRAM): $(OBJS) $(OBJS_DYNOPT) $(OTHEROBJS) $(GUEST_INIT)
+$(PROGRAM): $(OBJS) $(OBJS_DYNOPT) $(OTHEROBJS) $(GUEST_INIT) $(GUEST_PRE_INIT)
 	$(E) "  LINK    " $@
 	$(Q) $(CC) $(CFLAGS) $(OBJS) $(OBJS_DYNOPT) $(OTHEROBJS) $(GUEST_OBJS) $(LIBS) $(LIBS_DYNOPT) -o $@
 
@@ -372,9 +380,16 @@ $(PROGRAM_ALIAS): $(PROGRAM)
 	$(E) "  LN      " $@
 	$(Q) ln -f $(PROGRAM) $@
 
+ifneq ($(ARCH_PRE_INIT),)
+$(GUEST_PRE_INIT): $(ARCH_PRE_INIT)
+	$(E) "  LINK    " $@
+	$(Q) $(CC) -s -nostdlib $(ARCH_PRE_INIT) -o $@
+	$(Q) $(LD) $(LDFLAGS) -r -b binary -o guest/guest_pre_init.o $(GUEST_PRE_INIT)
+endif
+
 $(GUEST_INIT): guest/init.c
 	$(E) "  LINK    " $@
-	$(Q) $(CC) -static guest/init.c -o $@
+	$(Q) $(CC) $(GUEST_INIT_FLAGS) guest/init.c -o $@
 	$(Q) $(LD) $(LDFLAGS) -r -b binary -o guest/guest_init.o $(GUEST_INIT)
 
 %.s: %.c
@@ -473,7 +488,7 @@ clean:
 	$(Q) rm -f x86/bios/bios-rom.h
 	$(Q) rm -f tests/boot/boot_test.iso
 	$(Q) rm -rf tests/boot/rootfs/
-	$(Q) rm -f $(DEPS) $(OBJS) $(OTHEROBJS) $(OBJS_DYNOPT) $(STATIC_OBJS) $(PROGRAM) $(PROGRAM_ALIAS) $(PROGRAM)-static $(GUEST_INIT) $(GUEST_OBJS)
+	$(Q) rm -f $(DEPS) $(OBJS) $(OTHEROBJS) $(OBJS_DYNOPT) $(STATIC_OBJS) $(PROGRAM) $(PROGRAM_ALIAS) $(PROGRAM)-static $(GUEST_INIT) $(GUEST_PRE_INIT) $(GUEST_OBJS)
 	$(Q) rm -f cscope.*
 	$(Q) rm -f tags
 	$(Q) rm -f TAGS
-- 
2.4.3

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

[Oleg Nesterov]

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

Re: [PATCH 0/3] kvmtool: tiny init fox x86_64

[Pekka Enberg]

Hi Oleg,

On 10/19/2015 01:59 PM, Oleg Nesterov wrote:
> Yesterday I discovered kvmtool and it looks really cool! Thanks!
> Looks like, it does exactly what I need.
>
> But the static /virt/init doesn't look good. This series lessens
> the size of lkvm binary from 1045952 to 196200 bytes on x86_64,
> but this is minor. I want to write my /virt/init in shell or perl
> and this change can help.
>
> I don't really know who should be cc'ed, I've picked some names
> from git-log guest/init.c.

Will maintains the standalone kvmtool.git tree.

The series looks good to me:

Acked-by: Pekka Enberg <penberg@kernel.org>

- Pekka

Re: [PATCH 0/3] kvmtool: tiny init fox x86_64

[Oleg Nesterov]

Hi Pekka,

On 10/22, Pekka Enberg wrote:
>
> The series looks good to me:
>
> Acked-by: Pekka Enberg <penberg@kernel.org>

Thanks! I'll try to send some cleanups later.

But let me send another simple series first.

Oleg.