On Mon May 5, 2025 at 3:51 PM AEST, Akihiko Odaki wrote:
> On 2025/05/02 12:17, Nicholas Piggin wrote:
>> The msix interrupt throttling timer expiry sends an interrupt even if
>> there is no unmasked interrupt causes. This can be observed by seeing
>> two interrupts in response to a single event when throttling is active.
>>
>> The e1000e non-msix paths seem to get this right by masking and testing
>> ICR and IMS. Add similar checks for the msix cases in e1000e and igb.
>>
>> Signed-off-by: Nicholas Piggin <npig...@gmail.com>
>> ---
>> hw/net/e1000e_core.c | 10 ++++++----
>> hw/net/igb_core.c | 11 ++++++++---
>> 2 files changed, 14 insertions(+), 7 deletions(-)
>>
>> diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c
>> index 022884a2ea0..d53f70065ef 100644
>> --- a/hw/net/e1000e_core.c
>> +++ b/hw/net/e1000e_core.c
>> @@ -230,10 +230,12 @@ e1000e_intrmgr_on_msix_throttling_timer(void *opaque)
>>
>> timer->running = false;
>>
>> - causes = find_msix_causes(core, idx);
>> - trace_e1000e_irq_msix_notify_postponed_vec(idx);
>> - msix_notify(core->owner, idx);
>> - e1000e_msix_auto_clear_mask(core, causes);
>> + causes = find_msix_causes(core, idx) & core->mac[IMS] & core->mac[ICR];
>> + if (causes) {
>> + trace_e1000e_irq_msix_notify_postponed_vec(idx);
>> + msix_notify(core->owner, causes);
>
> I think this should be: msix_notify(core->owner, idx);
Yes good catch, it was fixed in the next patch but it's a but from
splitting. Thank you.
Thanks,
Nick
