Re: [PATCH v5 2/2] xen/evtchn: rework per event channel lock

[Jan Beulich <jbeulich@xxxxxxxx>]

On 09.11.2020 14:16, Jürgen Groß wrote:
> On 09.11.20 12:58, Jan Beulich wrote:
>> On 09.11.2020 07:41, Juergen Gross wrote:
>>> --- a/xen/arch/x86/pv/shim.c
>>> +++ b/xen/arch/x86/pv/shim.c
>>> @@ -660,11 +660,12 @@ void pv_shim_inject_evtchn(unsigned int port)
>>>       if ( port_is_valid(guest, port) )
>>>       {
>>>           struct evtchn *chn = evtchn_from_port(guest, port);
>>> -        unsigned long flags;
>>>   
>>> -        spin_lock_irqsave(&chn->lock, flags);
>>> -        evtchn_port_set_pending(guest, chn->notify_vcpu_id, chn);
>>> -        spin_unlock_irqrestore(&chn->lock, flags);
>>> +        if ( evtchn_read_trylock(chn) )
>>> +        {
>>> +            evtchn_port_set_pending(guest, chn->notify_vcpu_id, chn);
>>> +            evtchn_read_unlock(chn);
>>> +        }
>>
>> Does this need trylock?
> 
> It is called directly from the event upcall, so interrupts should be
> off here. Without trylock this would result in check_lock() triggering.

Hmm, right. Since the trylock function needs exposing anyway, this
isn't much of a problem, the more that it fits the pattern of being
a send.

Jan