Re: [Qemu-devel] [PATCH v2 19/23] blockdev: Drop DriveInfo member enable_auto_del

[Markus Armbruster]

Fam Zheng <address@hidden> writes:

> On Sat, 09/13 17:00, Markus Armbruster wrote:
>> Commit 2d246f0 introduced DriveInfo member enable_auto_del to
>> distinguish DriveInfo created via drive_new() from DriveInfo created
>> via qmp_blockdev_add().  The latter no longer exist.  Drop
>> enable_auto_del.
>> 
>> Signed-off-by: Markus Armbruster <address@hidden>
>> ---
>>  blockdev.c                | 11 +++--------
>>  include/sysemu/blockdev.h |  1 -
>>  2 files changed, 3 insertions(+), 9 deletions(-)
>> 
>> diff --git a/blockdev.c b/blockdev.c
>> index 0d99be0..e218c59 100644
>> --- a/blockdev.c
>> +++ b/blockdev.c
>> @@ -90,16 +90,14 @@ void blockdev_mark_auto_del(BlockBackend *blk)
>>      DriveInfo *dinfo = blk_legacy_dinfo(blk);
>>      BlockDriverState *bs = blk_bs(blk);
>>  
>> -    if (dinfo && !dinfo->enable_auto_del) {
>> +    if (!dinfo) {
>>          return;
>>      }
>>  
>>      if (bs->job) {
>>          block_job_cancel(bs->job);
>>      }
>> -    if (dinfo) {
>> -        dinfo->auto_del = 1;
>> -    }
>> +    dinfo->auto_del = 1;
>>  }
>
> Looks good.
>
> Based on the explanation in the commit message of patch 18, previouly,
> dinfo is always non-NULL, so enable_auto_del distinguishes whether the
> device is created by drive_new() or qmp_blockdev_add().

Correct.

> Now dinfo is NULL iff created by qmp_blockdev_add(), so
> enable_auto_del is not necessary any more.

Correct again.

> If I'm reading correctly, the dropped two "if (dinfo.."'s are also redundant.

Yes.  Let me explain why.

The two cases to distinguish are "created by drive_new()" and "created
by qmp_blockdev_add()".

Commit 2d246f0 introduced DriveInfo member enable_auto_del to exactly
that.

The code tests for "created by qmp_blockdev_add()" like this:

    dinfo && !dinfo->enable_auto_del

Before PATCH 18, this was actually unnecessarily careful, because dinfo
could never be null, but that's okay.

PATCH 18 changes the "created by qmp_blockdev_add() case: now dinfo is
null there.  The test remains correct, only now its second rather than
first part is unnecessary: dinfo now implies dinfo->enable_auto_del.

Before PATCH 18         dinfo != NULL   !dinfo->enable_auto_del   &&
by drive_new()          true            false                     false
by qmp_blockdev_add()   true            true                      true

After PATCH 18          dinfo != NULL   !dinfo->enable_auto_del   &&
by drive_new()          true            false                     false
by qmp_blockdev_add()   false           N/A                       false

Oops.

After PATCH 19          dinfo                                     !dinfo
by drive_new()          true                                      false
by qmp_blockdev_add()   false                                     true

Looks like I need to squash PATCH 19 into PATCH 18...  Thanks for making
me think!

[...]