From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:49179) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1dp9Tb-00039i-NX for qemu-devel@nongnu.org; Tue, 05 Sep 2017 04:45:00 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1dp9TS-0000ec-Is for qemu-devel@nongnu.org; Tue, 05 Sep 2017 04:44:51 -0400 Received: from mx0b-001b2d01.pphosted.com ([148.163.158.5]:53458 helo=mx0a-001b2d01.pphosted.com) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1dp9TS-0000dy-DT for qemu-devel@nongnu.org; Tue, 05 Sep 2017 04:44:46 -0400 Received: from pps.filterd (m0098417.ppops.net [127.0.0.1]) by mx0a-001b2d01.pphosted.com (8.16.0.21/8.16.0.21) with SMTP id v858dTaR040567 for ; Tue, 5 Sep 2017 04:44:45 -0400 Received: from e15.ny.us.ibm.com (e15.ny.us.ibm.com [129.33.205.205]) by mx0a-001b2d01.pphosted.com with ESMTP id 2csq86nd73-1 (version=TLSv1.2 cipher=AES256-SHA bits=256 verify=NOT) for ; Tue, 05 Sep 2017 04:44:45 -0400 Received: from localhost by e15.ny.us.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Tue, 5 Sep 2017 04:44:44 -0400 References: <1504239778-29893-1-git-send-email-zyimin@linux.vnet.ibm.com> <1504239778-29893-2-git-send-email-zyimin@linux.vnet.ibm.com> <20170905102928.6b23a28a.cohuck@redhat.com> From: Yi Min Zhao Date: Tue, 5 Sep 2017 16:44:37 +0800 MIME-Version: 1.0 In-Reply-To: <20170905102928.6b23a28a.cohuck@redhat.com> Content-Type: text/plain; charset=gbk; format=flowed Message-Id: <0923f97e-4f69-0bf3-b9ab-8e5f69a56bbb@linux.vnet.ibm.com> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH v2 1/3] s390x/pci: remove idx from msix msg data List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Cornelia Huck Cc: qemu-devel@nongnu.org, borntraeger@de.ibm.com, pasic@linux.vnet.ibm.com, pmorel@linux.vnet.ibm.com, agraf@suse.de, richard.henderson@linaro.org =D4=DA 2017/9/5 =CF=C2=CE=E74:29, Cornelia Huck =D0=B4=B5=C0: > On Fri, 1 Sep 2017 06:22:56 +0200 > Yi Min Zhao wrote: > >> PCIDevice pointer has been a parameter of kvm_arch_fixup_msi_route(). >> So we don't need to store zpci idx in msix message data to find out th= e >> specific zpci device. Instead, we could use pci device id to find its >> corresponding zpci device. >> >> Signed-off-by: Yi Min Zhao >> --- >> hw/s390x/s390-pci-bus.c | 16 +++++----------- >> hw/s390x/s390-pci-bus.h | 2 ++ >> hw/s390x/s390-pci-inst.c | 24 ------------------------ >> hw/s390x/s390-pci-stub.c | 6 ++++++ >> target/s390x/kvm.c | 7 +++++-- >> 5 files changed, 18 insertions(+), 37 deletions(-) >> >> diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c >> index 0a31a4ae88..bd8a3e1e1c 100644 >> --- a/hw/s390x/s390-pci-bus.c >> +++ b/hw/s390x/s390-pci-bus.c >> @@ -199,8 +199,8 @@ static S390PCIBusDevice *s390_pci_find_dev_by_uid(= S390pciState *s, uint16_t uid) >> return NULL; >> } >> =20 >> -static S390PCIBusDevice *s390_pci_find_dev_by_target(S390pciState *s, >> - const char *targ= et) >> +S390PCIBusDevice *s390_pci_find_dev_by_target(S390pciState *s, >> + const char *target) >> { >> S390PCIBusDevice *pbdev; >> =20 >> @@ -465,19 +465,13 @@ static void s390_msi_ctrl_write(void *opaque, hw= addr addr, uint64_t data, >> unsigned int size) >> { >> S390PCIBusDevice *pbdev =3D opaque; >> - uint32_t idx =3D data >> ZPCI_MSI_VEC_BITS; >> uint32_t vec =3D data & ZPCI_MSI_VEC_MASK; >> uint64_t ind_bit; >> uint32_t sum_bit; >> - uint32_t e =3D 0; >> =20 >> - DPRINTF("write_msix data 0x%" PRIx64 " idx %d vec 0x%x\n", data, = idx, vec); >> - >> - if (!pbdev) { >> - e |=3D (vec << ERR_EVENT_MVN_OFFSET); >> - s390_pci_generate_error_event(ERR_EVENT_NOMSI, idx, 0, addr, = e); >> - return; >> - } >> + assert(pbdev); > I'm wondering whether you could/should generate an error event here. > The one above probably won't work (as it seems to take idx as a > parameter), but is this really 'this must not happen, we messed up in > our code'? (Probably yes, but I want to be sure.) I think this must not happen. One a pci device is plugged into zPCI bus. We would assign a new memory region with zpci device as opaque for its msix. So if s390_msi_ctrl_write() is called, there must be a writ= e operation to a pci device's msix ctrl memory region which must has zpci device as a opaque. The construct is one-msi-mr-per-pci-device. > >> + DPRINTF("write_msix data 0x%" PRIx64 " idx %d vec 0x%x\n", data, >> + pbdev->idx, vec); >> =20 >> if (pbdev->state !=3D ZPCI_FS_ENABLED) { >> return; >> diff --git a/hw/s390x/s390-pci-stub.c b/hw/s390x/s390-pci-stub.c >> index 7a642d376c..e501e1b9ea 100644 >> --- a/hw/s390x/s390-pci-stub.c >> +++ b/hw/s390x/s390-pci-stub.c >> @@ -74,3 +74,9 @@ S390PCIBusDevice *s390_pci_find_dev_by_idx(S390pciSt= ate *s, uint32_t idx) >> { >> return NULL; >> } > Please remove s390_pci_find_dev_by_idx() from the stubs file, as it is > not used outside of the conditionally-built pci code anymore. I'm confused. s390_pci_find_dev_by_idx() can be called in=20 kvm_arch_fixup_msi_route(). And kvm_arch_fixup_msi_route() can be called by kvm_irqchip_add_msi_route= (). As the code, I think s390_pci_find_dev_by_idx() might be called. Could=20 you please explain more? > >> + >> +S390PCIBusDevice *s390_pci_find_dev_by_target(S390pciState *s, >> + const char *target) >> +{ >> + return NULL; >> +} >> diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c >> index 1338c29528..3d490c5e4b 100644 >> --- a/target/s390x/kvm.c >> +++ b/target/s390x/kvm.c >> @@ -2533,10 +2533,13 @@ int kvm_arch_fixup_msi_route(struct kvm_irq_ro= uting_entry *route, >> uint64_t address, uint32_t data, PCIDev= ice *dev) >> { >> S390PCIBusDevice *pbdev; >> - uint32_t idx =3D data >> ZPCI_MSI_VEC_BITS; >> uint32_t vec =3D data & ZPCI_MSI_VEC_MASK; >> =20 >> - pbdev =3D s390_pci_find_dev_by_idx(s390_get_phb(), idx); >> + if (!dev) { >> + return -ENODEV; > Can this actually happen? I think this cannot happen. But I'm afraid that I miss something. So I added this to avoid NULL pointer. But from the code and my test, there has not been NULL pointer happened. > >> + } >> + >> + pbdev =3D s390_pci_find_dev_by_target(s390_get_phb(), DEVICE(dev)= ->id); >> if (!pbdev) { >> DPRINTF("add_msi_route no dev\n"); >> return -ENODEV; >