On 24.06.21 10:42, Julien Grall wrote:
> Hi Juergen,
> 
> On 24/06/2021 10:30, Juergen Gross wrote:
>> On 16.06.21 16:43, Julien Grall wrote:
>>> From: Julien Grall <jgrall@amazon.com>
>>>
>>> Currently, the restore code is considering the stream will contain at
>>> most one in-flight request per connection. In a follow-up changes, we
>>> will want to transfer multiple in-flight requests.
>>>
>>> The function read_state_buffered() is now extended to restore multiple
>>> in-flight request. Complete requests will be queued as delayed
>>> requests, if there a partial request (only the last one can) then it
>>> will used as the current in-flight request.
>>>
>>> Note that we want to bypass the quota check for delayed requests as
>>> the new Xenstore may have a lower limit.
>>>
>>> Lastly, there is no need to change the specification as there was
>>> no restriction on the number of in-flight requests preserved.
>>>
>>> Signed-off-by: Julien Grall <jgrall@amazon.com>
>>> ---
>>>   tools/xenstore/xenstored_core.c | 56 ++++++++++++++++++++++++++++-----
>>>   1 file changed, 48 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/tools/xenstore/xenstored_core.c 
>>> b/tools/xenstore/xenstored_core.c
>>> index a5084a5b173d..5b7ab7f74013 100644
>>> --- a/tools/xenstore/xenstored_core.c
>>> +++ b/tools/xenstore/xenstored_core.c
>>> @@ -1486,6 +1486,10 @@ static void process_message(struct connection 
>>> *conn, struct buffered_data *in)
>>>       enum xsd_sockmsg_type type = in->hdr.msg.type;
>>>       int ret;
>>> +    /* At least send_error() and send_reply() expects 
conn->in == in */
>>> +    assert(conn->in == in);
>>> +    trace_io(conn, in, 0);
>>> +
>>>       if ((unsigned int)type >= XS_TYPE_COUNT|| !wire_funcs[type].func) 
>> {
>>>           eprintf("Client unknown operation %i", type);
>>>           send_error(conn, ENOSYS);
>>> @@ -1515,6 +1519,23 @@ static void process_message(struct connection 
>>> *conn, struct buffered_data *in)
>>>       conn->transaction = NULL;
>>>   }
>>> +static bool process_delayed_message(struct delayed_request *req)
>>> +{
>>> +    struct connection *conn = req->data;
>>> +    struct buffered_data *saved_in = conn->in;
>>> +
>>> +    /*
>>> +     * Part of process_message() expects conn->in to contains the
>>> +     * processed response. So save the current conn->in and restore it
>>> +     * afterwards.
>>> +     */
>>> +    conn->in = req->in;
>>> +    process_message(req->data, req->in);
>>> +    conn->in = saved_in;
>>
>> This pattern was added to do_lu_start() already, and it will be needed
>> for each other function being called via call_delayed().
>>
>> Maybe it would be better to add conn explicitly to struct
>> delayed_request (or even better: replace data with conn) and to do the
>> conn->in saving/setting/restoring in call_delayed() instead?
> 
> This was my original approach, but I abandoned it because in the case of 
> do_lu_start() we need to save the original conn->in in the stream (see 
> patch #3 for more details).
> 
> If we overwrite conn->in in call_delayed(), then we need a different way 
> to find the original conn->in. I couldn't find a nice way and decided to 
> settle with the duplication.

Ah, okay, understood. Even in case we'd be able to restore conn->in it
would be the same amount of code again, but harder to follow.


Juergen