On Sun Jan 19, 2025 at 7:22 PM AEST, Yan Vugenfirer wrote: > On Fri, Jan 17, 2025 at 7:05 PM Nicholas Piggin <npig...@gmail.com> wrote: > >> Assert there is no existing irq raised that would lead to a false >> positive interrupt test. >> >> e1000e has to disable interrupt throttling for this test, because >> it can cause delayed superfluous interrupts which trip the assertions. >> >> Signed-off-by: Nicholas Piggin <npig...@gmail.com> >> --- >> tests/qtest/libqos/e1000e.h | 1 + >> tests/qtest/e1000e-test.c | 10 ++++++++++ >> tests/qtest/igb-test.c | 6 ++++++ >> tests/qtest/libqos/e1000e.c | 9 ++++++++- >> 4 files changed, 25 insertions(+), 1 deletion(-) >> >> diff --git a/tests/qtest/libqos/e1000e.h b/tests/qtest/libqos/e1000e.h >> index 30643c80949..7154aec0339 100644 >> --- a/tests/qtest/libqos/e1000e.h >> +++ b/tests/qtest/libqos/e1000e.h >> @@ -54,6 +54,7 @@ static inline uint32_t e1000e_macreg_read(QE1000E *d, >> uint32_t reg) >> return qpci_io_readl(&d_pci->pci_dev, d_pci->mac_regs, reg); >> } >> >> +bool e1000e_seen_isr(QE1000E *d, uint16_t msg_id); >> void e1000e_wait_isr(QE1000E *d, uint16_t msg_id); >> void e1000e_tx_ring_push(QE1000E *d, void *descr); >> void e1000e_rx_ring_push(QE1000E *d, void *descr); >> diff --git a/tests/qtest/e1000e-test.c b/tests/qtest/e1000e-test.c >> index 746d26cfb67..9ab81ecff5b 100644 >> --- a/tests/qtest/e1000e-test.c >> +++ b/tests/qtest/e1000e-test.c >> @@ -61,6 +61,9 @@ static void e1000e_send_verify(QE1000E *d, int >> *test_sockets, QGuestAllocator *a >> E1000_TXD_DTYP_D | >> sizeof(buffer)); >> >> + /* Ensure the interrupt has not been taken already */ >> + g_assert(!e1000e_seen_isr(d, E1000E_TX0_MSG_ID)); >> + >> /* Put descriptor to the ring */ >> e1000e_tx_ring_push(d, &descr); >> >> @@ -105,6 +108,9 @@ static void e1000e_receive_verify(QE1000E *d, int >> *test_sockets, QGuestAllocator >> char buffer[64]; >> int ret; >> >> + /* Ensure the interrupt has not been taken already */ >> + g_assert(!e1000e_seen_isr(d, E1000E_RX0_MSG_ID)); >> + >> > I don't think potentially crashing the guest is the right solution. > I suggest maybe substituting with logging.
This is just for qtest where assertions are used to indicate failure. Thanks, Nick
