Patchwork [+STABLE-0.14] exit if -drive specified is invalid instead of ignoring the "wrong" -drive

login
register
mail settings
Submitter Michael Tokarev
Date March 17, 2011, 2:28 p.m.
Message ID <4D821A7F.9000204@msgid.tls.msk.ru>
Download mbox | patch
Permalink /patch/87380/
State New
Headers show

Comments

Michael Tokarev - March 17, 2011, 2:28 p.m.
17.03.2011 16:51, Markus Armbruster wrote:
> Michael Tokarev <mjt@tls.msk.ru> writes:
> 
>> Trivial patch.  I've sent it yesterday but somehow it didn't
>> reach the list.
>>
>> This fixes the problem when qemu continues even if -drive specification
>> is somehow invalid, resulting in a mess.  Applicable for both current
>> master and for stable-0.14 (and 0.13 and 0.12 as well).
> 
> Note patch doesn't apply to 0.12 and 0.13.

Yes it doesn't, since to 0.14 the code changed.
What I mean is that the same problem exist in
earlier versions too.  I'll apply a change of
this effect to 0.12.5 as used in Debian now,
something like the one below.

[]
> What about all the other unchecked drive_add() calls in main()?

These are much less worrisome - they fail only of the
internal definitions of options are incorrect, which
should never happen.  For example:

            case QEMU_OPTION_hdd:
                drive_add(IF_DEFAULT, popt->index - QEMU_OPTION_hda, optarg,
                          HD_OPTS);

There, optarg is just a filename, and HD_OPTS is
defined like this:

#define HD_OPTS "media=disk"

So it should not fail when parsing options, only
when trying to interpret and actually open the
filename, which happens much later in the game.

Thanks,

/mjt

For 0.12 and 0.13:
Markus Armbruster - March 17, 2011, 3:35 p.m.
Michael Tokarev <mjt@tls.msk.ru> writes:

> 17.03.2011 16:51, Markus Armbruster wrote:
>> Michael Tokarev <mjt@tls.msk.ru> writes:
>> 
>>> Trivial patch.  I've sent it yesterday but somehow it didn't
>>> reach the list.
>>>
>>> This fixes the problem when qemu continues even if -drive specification
>>> is somehow invalid, resulting in a mess.  Applicable for both current
>>> master and for stable-0.14 (and 0.13 and 0.12 as well).
>> 
>> Note patch doesn't apply to 0.12 and 0.13.
>
> Yes it doesn't, since to 0.14 the code changed.
> What I mean is that the same problem exist in
> earlier versions too.  I'll apply a change of
> this effect to 0.12.5 as used in Debian now,
> something like the one below.
>
> []
>> What about all the other unchecked drive_add() calls in main()?
>
> These are much less worrisome - they fail only of the
> internal definitions of options are incorrect, which
> should never happen.  For example:
>
>             case QEMU_OPTION_hdd:
>                 drive_add(IF_DEFAULT, popt->index - QEMU_OPTION_hda, optarg,
>                           HD_OPTS);
>
> There, optarg is just a filename, and HD_OPTS is
> defined like this:
>
> #define HD_OPTS "media=disk"
>
> So it should not fail when parsing options, only
> when trying to interpret and actually open the
> filename, which happens much later in the game.

Fair enough.

Patch

diff --git a/vl.c b/vl.c
index 77677e8..069a1df 100644
--- a/vl.c
+++ b/vl.c
@@ -5025,7 +5025,8 @@  int main(int argc, char **argv, char **envp)
                 drive_add(optarg, HD_ALIAS, popt->index - QEMU_OPTION_hda);
                 break;
             case QEMU_OPTION_drive:
-                drive_add(NULL, "%s", optarg);
+                if (drive_add(NULL, "%s", optarg) == NULL)
+                    exit(1);
                break;
             case QEMU_OPTION_set:
                 if (qemu_set_option(optarg) != 0)