Re: [Qemu-devel] [PATCH] don't exit if cdrom media fails to open

[Cole Robinson <crobinso@redhat.com>]

David Ahern wrote:
> 
> Mark McLoughlin wrote:
>> Hi David,
>>
>> On Tue, 2009-06-09 at 07:41 -0600, David Ahern wrote:
>>> This patch allows a VM to continue even if the cdrom image is not
>>> accessible at VM boot time. It allows a boot progression similar to real
>>> hardware (e.g., try cd, then disk).
>>>
>>> For example,
>>>
>>> qemu-system-x86_64 -cdrom /dev/dvd -hda disk.img -boot dc
>>>
>>> If there is no media in the tray, qemu currently exits with the message:
>>> qemu: could not open disk image /dev/dvd
>>>
>>> With this patch, the message is still displayed, but qemu continues. It
>>> first tries to boot from the cdrom and then falls back to the disk.
>>>
>>> Signed-off-by: David Ahern <dsahern@gmail.com>
>>>
>>> ---
>>>
>>> diff --git a/vl.c b/vl.c
>>> index fcf8532..bd3709c 100644
>>> --- a/vl.c
>>> +++ b/vl.c
>>> @@ -152,6 +152,7 @@ int main(int argc, char **argv)
>>>  #include "qemu-char.h"
>>>  #include "cache-utils.h"
>>>  #include "block.h"
>>> +#include "block_int.h"
>>>  #include "dma.h"
>>>  #include "audio/audio.h"
>>>  #include "migration.h"
>>> @@ -2551,7 +2552,8 @@ int drive_init(struct drive_opt *arg, int
>>> snapshot, void *opaque)
>>>      if (bdrv_open2(bdrv, file, bdrv_flags, drv) < 0) {
>>>          fprintf(stderr, "qemu: could not open disk image %s\n",
>>>                          file);
>>> -        return -1;
>>> +        if (bdrv->type != BDRV_TYPE_CDROM)
>>> +            return -1;
>> I think Cole's patch here:
>>
>>   https://bugzilla.redhat.com/show_bug.cgi?id=473154#c15
>>
>> fixes the same problem? Certainly, it seems to me that any code
>> requiring the path to be e.g. /dev/cd is doomed and should be fixed as
>> Cole suggests?
> 
> Cole's patch works as well. I don't have a preference for how it gets
> done, just that I can start a guest without checking whether there is
> media in the tray and adjusting the qemu invocation.
> 

There is an extra benefit to my patch: /dev/dvd will be treated as a
host cdrom device, so ejecting media in the guest will be mapped to a
physical eject on the host.

I'll touch up my patch and send with a Signed-off-by.

Thanks,
Cole