[pve-devel] [PATCH qemu 1/2] PVE: Don't expect complete_cb to be called outside coroutine

[Stefan Reiter]

On 10/27/20 3:16 PM, Wolfgang Bumiller wrote:
> On Thu, Oct 22, 2020 at 02:11:17PM +0200, Stefan Reiter wrote:
>> We're at the mercy of the rest of QEMU here, and it sometimes decides to
>> call pvebackup_complete_cb from a coroutine. This really doesn't matter
>> to us, so don't assert and crash on it.
>>
>> Signed-off-by: Stefan Reiter <s.reiter at proxmox.com>
>> ---
>>   pve-backup.c | 7 +++----
>>   1 file changed, 3 insertions(+), 4 deletions(-)
>>
>> diff --git a/pve-backup.c b/pve-backup.c
>> index 53cf23ed5a..9179754dcb 100644
>> --- a/pve-backup.c
>> +++ b/pve-backup.c
>> @@ -318,19 +318,18 @@ static void coroutine_fn pvebackup_co_complete_stream(void *opaque)
>>   
>>   static void pvebackup_complete_cb(void *opaque, int ret)
>>   {
>> -    assert(!qemu_in_coroutine());
>> -
>>       PVEBackupDevInfo *di = opaque;
>>       di->completed_ret = ret;
>>   
>>       /*
>>        * Schedule stream cleanup in async coroutine. close_image and finish might
>> -     * take a while, so we can't block on them here.
>> +     * take a while, so we can't block on them here. This way it also doesn't
>> +     * matter if we're already running in a coroutine or not.
>>        * Note: di is a pointer to an entry in the global backup_state struct, so
>>        * it stays valid.
>>        */
>>       Coroutine *co = qemu_coroutine_create(pvebackup_co_complete_stream, di);
>> -    aio_co_schedule(qemu_get_aio_context(), co);
>> +    aio_co_enter(qemu_get_aio_context(), co);
> 
> Shouldn't this be decided based on `qemu_in_coroutine()`? Or are we
> allowed to call enter regardless, I forgot...?
> 

We are allowed to call whenever, if we're in a coroutine in the same 
context this becomes a call, if not it's scheduled correctly.

>>   }
>>   
>>   static void coroutine_fn pvebackup_co_cancel(void *opaque)
>> -- 
>> 2.20.1