Received: from eggs.gnu.org ([2001:4830:134:3::10]:47989) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1f2kMS-0001OF-9S for qemu-devel@nongnu.org; Sun, 01 Apr 2018 17:18:01 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1f2kMP-00061G-3t for qemu-devel@nongnu.org; Sun, 01 Apr 2018 17:18:00 -0400 Received: from mout.web.de ([212.227.15.14]:57529) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1f2kMO-00060k-Pj for qemu-devel@nongnu.org; Sun, 01 Apr 2018 17:17:57 -0400 From: Jan Kiszka Message-ID: <79199357-38dc-4a2a-6e67-e80235448b31@web.de> Date: Sun, 1 Apr 2018 23:17:53 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit Subject: [Qemu-devel] [PATCH] e1000e: Prevent MSI/MSI-X storms List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Dmitry Fleytman , qemu-devel From: Jan Kiszka Only signal MSI/MSI-X events on rising edges. So far we re-triggered the interrupt sources even if the guest did no consumed the pending one, easily causing interrupt storms. Issue was observable with Linux 4.16 e1000e driver when MSI-X was used. Vector 2 was causing interrupt storms after the driver activated the device. Signed-off-by: Jan Kiszka --- hw/net/e1000e_core.c | 10 ++++++++++ hw/net/e1000e_core.h | 2 ++ 2 files changed, 12 insertions(+) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index c93c4661ed..ecf9b15555 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -2123,6 +2123,13 @@ e1000e_send_msi(E1000ECore *core, bool msix) { uint32_t causes = core->mac[ICR] & core->mac[IMS] & ~E1000_ICR_ASSERTED; + core->msi_causes_pending &= causes; + causes ^= core->msi_causes_pending; + if (causes == 0) { + return; + } + core->msi_causes_pending |= causes; + if (msix) { e1000e_msix_notify(core, causes); } else { @@ -2160,6 +2167,9 @@ e1000e_update_interrupt_state(E1000ECore *core) core->mac[ICS] = core->mac[ICR]; interrupts_pending = (core->mac[IMS] & core->mac[ICR]) ? true : false; + if (!interrupts_pending) { + core->msi_causes_pending = 0; + } trace_e1000e_irq_pending_interrupts(core->mac[ICR] & core->mac[IMS], core->mac[ICR], core->mac[IMS]); diff --git a/hw/net/e1000e_core.h b/hw/net/e1000e_core.h index 7d8ff41890..63a15510cc 100644 --- a/hw/net/e1000e_core.h +++ b/hw/net/e1000e_core.h @@ -109,6 +109,8 @@ struct E1000Core { NICState *owner_nic; PCIDevice *owner; void (*owner_start_recv)(PCIDevice *d); + + uint32_t msi_causes_pending; }; void -- 2.13.6