[Qemu-devel] [PATCH] hw/i386/pc: reject to boot a wrong header magic kernel

[Qemu-devel] [PATCH] hw/i386/pc: reject to boot a wrong header magic kernel

[liguang]

if head magic is missing or wrong unexpectedly, we'd
better to reject booting.
e.g.
I make a mistake to boot a vmlinuz for MIPS(which
I think it's for x86) like this:
qemu-system-x86_64 -kernel vmlinuz -initrd demord
then qemu report:
"qemu: linux kernel too old to load a ram disk"
that's misleading.

Signed-off-by: liguang <lig.fnst@cn.fujitsu.com>
---
 hw/i386/pc.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/hw/i386/pc.c b/hw/i386/pc.c
index b1e06fa..2b78dfc 100644
--- a/hw/i386/pc.c
+++ b/hw/i386/pc.c
@@ -683,8 +683,10 @@ static void load_linux(void *fw_cfg,
         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
                            kernel_cmdline, kernel_size, header)) {
             return;
+        } else {
+            fprintf(stderr, "please assure specicified kernel is for x86!\n");
+            exit(1);
         }
-        protocol = 0;
     }
 
     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
-- 
1.7.2.5

Re: [Qemu-devel] [Qemu-trivial] [PATCH] hw/i386/pc: reject to boot a wrong header magic kernel

[Stefan Hajnoczi]

On Wed, Mar 27, 2013 at 02:10:31PM +0800, liguang wrote:
> if head magic is missing or wrong unexpectedly, we'd
> better to reject booting.
> e.g.
> I make a mistake to boot a vmlinuz for MIPS(which
> I think it's for x86) like this:
> qemu-system-x86_64 -kernel vmlinuz -initrd demord
> then qemu report:
> "qemu: linux kernel too old to load a ram disk"
> that's misleading.
> 
> Signed-off-by: liguang <lig.fnst@cn.fujitsu.com>
> ---
>  hw/i386/pc.c |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> index b1e06fa..2b78dfc 100644
> --- a/hw/i386/pc.c
> +++ b/hw/i386/pc.c
> @@ -683,8 +683,10 @@ static void load_linux(void *fw_cfg,
>          if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
>                             kernel_cmdline, kernel_size, header)) {
>              return;
> +        } else {
> +            fprintf(stderr, "please assure specicified kernel is for x86!\n");
> +            exit(1);

load_multiboot() can fail for other reasons so this error messing is
misleading.  Giving QEMU a non-x86 kernel is just one scenario where
this may fail.

>          }
> -        protocol = 0;
>      }

Why did you drop protocol = 0?

Stefan

Re: [Qemu-devel] [Qemu-trivial] [PATCH] hw/i386/pc: reject to boot a wrong header magic kernel

[li guang]

在 2013-03-28四的 10:42 +0100，Stefan Hajnoczi写道：
> On Wed, Mar 27, 2013 at 02:10:31PM +0800, liguang wrote:
> > if head magic is missing or wrong unexpectedly, we'd
> > better to reject booting.
> > e.g.
> > I make a mistake to boot a vmlinuz for MIPS(which
> > I think it's for x86) like this:
> > qemu-system-x86_64 -kernel vmlinuz -initrd demord
> > then qemu report:
> > "qemu: linux kernel too old to load a ram disk"
> > that's misleading.
> > 
> > Signed-off-by: liguang <lig.fnst@cn.fujitsu.com>
> > ---
> >  hw/i386/pc.c |    4 +++-
> >  1 files changed, 3 insertions(+), 1 deletions(-)
> > 
> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> > index b1e06fa..2b78dfc 100644
> > --- a/hw/i386/pc.c
> > +++ b/hw/i386/pc.c
> > @@ -683,8 +683,10 @@ static void load_linux(void *fw_cfg,
> >          if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
> >                             kernel_cmdline, kernel_size, header)) {
> >              return;
> > +        } else {
> > +            fprintf(stderr, "please assure specicified kernel is for x86!\n");
> > +            exit(1);
> 
> load_multiboot() can fail for other reasons so this error messing is
> misleading.  Giving QEMU a non-x86 kernel is just one scenario where
> this may fail.

according to my check of load_mutiboot function,
mostly it will return 0 if it's not multboot,
or 1 it's a multiboot, so print this message,
or can I just print "wrong kernel image!" ?

> 
> >          }
> > -        protocol = 0;
> >      }
> 
> Why did you drop protocol = 0?

I think we only want either normal or multi boot linux kernel,
I can't see meaning let other case go on.
so, here, if a normal kernel, OK, go on,
if multiboot, OK, go on,
others, NO, end up.
so, "protocol = 0" is meaningless here.

Re: [Qemu-devel] [Qemu-trivial] [PATCH] hw/i386/pc: reject to boot a wrong header magic kernel

[Stefan Hajnoczi]

On Fri, Mar 29, 2013 at 1:38 AM, li guang <lig.fnst@cn.fujitsu.com> wrote:
> 在 2013-03-28四的 10:42 +0100，Stefan Hajnoczi写道：
>> On Wed, Mar 27, 2013 at 02:10:31PM +0800, liguang wrote:
>> > if head magic is missing or wrong unexpectedly, we'd
>> > better to reject booting.
>> > e.g.
>> > I make a mistake to boot a vmlinuz for MIPS(which
>> > I think it's for x86) like this:
>> > qemu-system-x86_64 -kernel vmlinuz -initrd demord
>> > then qemu report:
>> > "qemu: linux kernel too old to load a ram disk"
>> > that's misleading.
>> >
>> > Signed-off-by: liguang <lig.fnst@cn.fujitsu.com>
>> > ---
>> >  hw/i386/pc.c |    4 +++-
>> >  1 files changed, 3 insertions(+), 1 deletions(-)
>> >
>> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
>> > index b1e06fa..2b78dfc 100644
>> > --- a/hw/i386/pc.c
>> > +++ b/hw/i386/pc.c
>> > @@ -683,8 +683,10 @@ static void load_linux(void *fw_cfg,
>> >          if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
>> >                             kernel_cmdline, kernel_size, header)) {
>> >              return;
>> > +        } else {
>> > +            fprintf(stderr, "please assure specicified kernel is for x86!\n");
>> > +            exit(1);
>>
>> load_multiboot() can fail for other reasons so this error messing is
>> misleading.  Giving QEMU a non-x86 kernel is just one scenario where
>> this may fail.
>
> according to my check of load_mutiboot function,
> mostly it will return 0 if it's not multboot,
> or 1 it's a multiboot, so print this message,
> or can I just print "wrong kernel image!" ?

Yes, load_multiboot() fails if the image is not a valid multiboot
image.  An error message like "not a valid multiboot image" is good.

>>
>> >          }
>> > -        protocol = 0;
>> >      }
>>
>> Why did you drop protocol = 0?
>
> I think we only want either normal or multi boot linux kernel,
> I can't see meaning let other case go on.
> so, here, if a normal kernel, OK, go on,
> if multiboot, OK, go on,
> others, NO, end up.
> so, "protocol = 0" is meaningless here.

I looked in more detail now and here is why we cannot drop this line of code:

    uint16_t protocol;

Note that the variable is not initialized.

[...]
    if (ldl_p(header+0x202) == 0x53726448)
    protocol = lduw_p(header+0x206);
    else {
    /* This looks like a multiboot kernel. If it is, let's stop
       treating it like a Linux kernel. */
        if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
                           kernel_cmdline, kernel_size, header))
            return;
    protocol = 0;

Set it to 0 here so it's initialized

    }
    if (protocol < 0x200 || !(header[0x211] & 0x01)) {

Use variable here - so we *must* initialize it before use.

Stefan

Re: [Qemu-devel] [Qemu-trivial] [PATCH] hw/i386/pc: reject to boot a wrong header magic kernel

[li guang]

OK, Thanks!
I will update this patch as your comment.

在 2013-03-29五的 15:46 +0100，Stefan Hajnoczi写道：
> On Fri, Mar 29, 2013 at 1:38 AM, li guang <lig.fnst@cn.fujitsu.com> wrote:
> > 在 2013-03-28四的 10:42 +0100，Stefan Hajnoczi写道：
> >> On Wed, Mar 27, 2013 at 02:10:31PM +0800, liguang wrote:
> >> > if head magic is missing or wrong unexpectedly, we'd
> >> > better to reject booting.
> >> > e.g.
> >> > I make a mistake to boot a vmlinuz for MIPS(which
> >> > I think it's for x86) like this:
> >> > qemu-system-x86_64 -kernel vmlinuz -initrd demord
> >> > then qemu report:
> >> > "qemu: linux kernel too old to load a ram disk"
> >> > that's misleading.
> >> >
> >> > Signed-off-by: liguang <lig.fnst@cn.fujitsu.com>
> >> > ---
> >> >  hw/i386/pc.c |    4 +++-
> >> >  1 files changed, 3 insertions(+), 1 deletions(-)
> >> >
> >> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> >> > index b1e06fa..2b78dfc 100644
> >> > --- a/hw/i386/pc.c
> >> > +++ b/hw/i386/pc.c
> >> > @@ -683,8 +683,10 @@ static void load_linux(void *fw_cfg,
> >> >          if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
> >> >                             kernel_cmdline, kernel_size, header)) {
> >> >              return;
> >> > +        } else {
> >> > +            fprintf(stderr, "please assure specicified kernel is for x86!\n");
> >> > +            exit(1);
> >>
> >> load_multiboot() can fail for other reasons so this error messing is
> >> misleading.  Giving QEMU a non-x86 kernel is just one scenario where
> >> this may fail.
> >
> > according to my check of load_mutiboot function,
> > mostly it will return 0 if it's not multboot,
> > or 1 it's a multiboot, so print this message,
> > or can I just print "wrong kernel image!" ?
> 
> Yes, load_multiboot() fails if the image is not a valid multiboot
> image.  An error message like "not a valid multiboot image" is good.
> 
> >>
> >> >          }
> >> > -        protocol = 0;
> >> >      }
> >>
> >> Why did you drop protocol = 0?
> >
> > I think we only want either normal or multi boot linux kernel,
> > I can't see meaning let other case go on.
> > so, here, if a normal kernel, OK, go on,
> > if multiboot, OK, go on,
> > others, NO, end up.
> > so, "protocol = 0" is meaningless here.
> 
> I looked in more detail now and here is why we cannot drop this line of code:
> 
>     uint16_t protocol;
> 
> Note that the variable is not initialized.
> 
> [...]
>     if (ldl_p(header+0x202) == 0x53726448)
>     protocol = lduw_p(header+0x206);
>     else {
>     /* This looks like a multiboot kernel. If it is, let's stop
>        treating it like a Linux kernel. */
>         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
>                            kernel_cmdline, kernel_size, header))
>             return;
>     protocol = 0;
> 
> Set it to 0 here so it's initialized
> 
>     }
>     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
> 
> Use variable here - so we *must* initialize it before use.
> 
> Stefan