[libvirt] [PATCH] qemu: fix crash in migrate when migrateuri do not have a scheme

[lhuang]

On 02/11/2015 04:19 PM, Ján Tomko wrote:
> On Wed, Feb 11, 2015 at 03:50:37PM +0800, Shanzhi Yu wrote:
>> On 02/11/2015 03:41 PM, Luyao Huang wrote:
>>> https://bugzilla.redhat.com/show_bug.cgi?id=1191355
>>>
>>> When we migrate a vm with migrateuri option with a uri do not
>>> have scheme like this:
>>>
>>>    # virsh migrate test4 --live qemu+ssh://lhuang/system --migrateuri 127.0.0.1
>>>
>>> target libvirtd will crashed because uri->scheme is NULL in
>>> qemuMigrationPrepareDirect this line:
>>>
>>>        if (STRNEQ(uri->scheme, "tcp") &&
> There is a similar check in doNativeMigrate:
>
>     if (!(uribits = qemuMigrationParseURI(uri, NULL)))
>          return -1;
>
>      if (STREQ(uribits->scheme, "rdma")) {
>          if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_RDMA)) {
>              virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
>                             _("outgoing RDMA migration is not supported "
>                               "with this QEMU binary"));
>
> It seems the scheme can be NULL here only if Prepare on the remote side
> returned a wrong URI, It would still be nice not to crash in that case.
>

Thanks for pointing out and i will send a v2 later.
>>> add a value check before this line.
>>>
>>> Signed-off-by: Luyao Huang <lhuang at redhat.com>
>>> ---
>>>    src/qemu/qemu_migration.c | 7 +++++++
>>>    1 file changed, 7 insertions(+)
>>>
>>> diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c
>>> index 879b1bf..5c3b73e 100644
>>> --- a/src/qemu/qemu_migration.c
>>> +++ b/src/qemu/qemu_migration.c
>>> @@ -3281,6 +3281,13 @@ qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
>>>            if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
>>>                goto cleanup;
>>>    
>>> +        if (uri->scheme == NULL) {
>>> +            virReportError(VIR_ERR_INVALID_ARG,
>>> +                           _("missing scheme in migration URI: %s"),
>>> +                           uri_in);
>>> +            goto cleanup;
>>> +        }
>>> +
>>>            if (STRNEQ(uri->scheme, "tcp") &&
>>>                STRNEQ(uri->scheme, "rdma")) {
>> Why not just use "STRNEQ_NULLABLE" instead of "STRNEQ" directly?
>>
> It would report 'unsupported scheme (null) in migration URI:',
> instead of saying that the scheme is missing.
>
> Jan

Luyao