When a PF is bound to the driver via driver_override and passed through to a VM, its pf_qm stays NULL. The PCI error handler reset_prepare() runs during open_device through pci_try_reset_function(), before the mig_ops gate, and dereferences the NULL pf_qm for the timeout log, crashing the kernel. Move the mig_ops check to the entry of reset_prepare() and aer_reset_done() so non-migration devices skip the QM_RESETTING coordination. The dev local resolved &qm->pdev->dev before the gate, so open code it at the dev_err() call where qm is known valid. Also clear set_reset_flag together with QM_RESETTING in aer_reset_done(); the flag was never cleared before, so a later timed-out reset could release a foreign lock. Fixes: b0eed085903e ("hisi_acc_vfio_pci: Add support for VFIO live migration") Fixes: a22099ed7936f ("hisi_acc_vfio_pci: fix VF reset timeout issue") Signed-off-by: Longfang Liu --- drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-) diff --git a/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c b/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c index 86362ec424a5..e6fdf4d2e581 100644 --- a/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +++ b/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c @@ -1154,14 +1154,16 @@ static void hisi_acc_vf_pci_reset_prepare(struct pci_dev *pdev) { struct hisi_acc_vf_core_device *hisi_acc_vdev = hisi_acc_drvdata(pdev); struct hisi_qm *qm = hisi_acc_vdev->pf_qm; - struct device *dev = &qm->pdev->dev; u32 delay = 0; + if (!hisi_acc_vdev->core_device.vdev.mig_ops) + return; + /* All reset requests need to be queued for processing */ while (test_and_set_bit(QM_RESETTING, &qm->misc_ctl)) { msleep(1); if (++delay > QM_RESET_WAIT_TIMEOUT) { - dev_err(dev, "reset prepare failed\n"); + dev_err(&qm->pdev->dev, "reset prepare failed\n"); return; } } @@ -1174,12 +1176,14 @@ static void hisi_acc_vf_pci_aer_reset_done(struct pci_dev *pdev) struct hisi_acc_vf_core_device *hisi_acc_vdev = hisi_acc_drvdata(pdev); struct hisi_qm *qm = hisi_acc_vdev->pf_qm; - if (hisi_acc_vdev->set_reset_flag) - clear_bit(QM_RESETTING, &qm->misc_ctl); - if (!hisi_acc_vdev->core_device.vdev.mig_ops) return; + if (hisi_acc_vdev->set_reset_flag) { + clear_bit(QM_RESETTING, &qm->misc_ctl); + hisi_acc_vdev->set_reset_flag = false; + } + mutex_lock(&hisi_acc_vdev->state_mutex); hisi_acc_vf_reset(hisi_acc_vdev); mutex_unlock(&hisi_acc_vdev->state_mutex); -- 2.43.0 On QM_HW_V3 (VF_CTRL mode) hardware with 64KB pages, the functional region and migration registers share one BAR2 physical page with no isolation, which leaks migration registers to the guest and causes guest kernel calltrace under the KVM emulated device scheme. The previous fix rejected the device in hisi_acc_vf_qm_init() during open_device, returning -EINVAL and breaking passthrough entirely. Move the drv_mode check into a helper called from probe() so that a VF_CTRL device on a 64KB page is bound to the generic ops, keeping passthrough working while disabling live migration. Fixes: b0eed085903e ("hisi_acc_vfio_pci: Add support for VFIO live migration") Signed-off-by: Longfang Liu --- .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 32 ++++++++++++------- 1 file changed, 20 insertions(+), 12 deletions(-) diff --git a/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c b/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c index e6fdf4d2e581..818ad5112d7b 100644 --- a/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +++ b/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c @@ -1195,13 +1195,6 @@ static int hisi_acc_vf_qm_init(struct hisi_acc_vf_core_device *hisi_acc_vdev) struct hisi_qm *vf_qm = &hisi_acc_vdev->vf_qm; struct hisi_qm *pf_qm = hisi_acc_vdev->pf_qm; struct pci_dev *vf_dev = vdev->pdev; - u32 val; - - val = readl(pf_qm->io_base + QM_MIG_REGION_SEL); - if (pf_qm->ver > QM_HW_V3 && (val & QM_MIG_REGION_EN)) - hisi_acc_vdev->drv_mode = HW_ACC_MIG_PF_CTRL; - else - hisi_acc_vdev->drv_mode = HW_ACC_MIG_VF_CTRL; if (hisi_acc_vdev->drv_mode == HW_ACC_MIG_PF_CTRL) { /* @@ -1669,21 +1662,34 @@ static void hisi_acc_vf_debugfs_exit(struct hisi_acc_vf_core_device *hisi_acc_vd hisi_acc_vdev->debug_migf = NULL; } +static enum hw_drv_mode hisi_acc_vf_get_drv_mode(struct hisi_qm *pf_qm) +{ + u32 val; + + val = readl(pf_qm->io_base + QM_MIG_REGION_SEL); + if (pf_qm->ver > QM_HW_V3 && (val & QM_MIG_REGION_EN)) + return HW_ACC_MIG_PF_CTRL; + + return HW_ACC_MIG_VF_CTRL; +} + static int hisi_acc_vfio_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id) { struct hisi_acc_vf_core_device *hisi_acc_vdev; const struct vfio_device_ops *ops = &hisi_acc_vfio_pci_ops; + enum hw_drv_mode drv_mode = HW_ACC_MIG_VF_CTRL; + resource_size_t func_len; struct hisi_qm *pf_qm; - int vf_id; int ret; pf_qm = hisi_acc_get_pf_qm(pdev); if (pf_qm && pf_qm->ver >= QM_HW_V3) { - vf_id = pci_iov_vf_id(pdev); - if (vf_id >= 0) - ops = &hisi_acc_vfio_pci_migrn_ops; + func_len = pci_resource_len(pdev, VFIO_PCI_BAR2_REGION_INDEX) >> 1; + drv_mode = hisi_acc_vf_get_drv_mode(pf_qm); + if (drv_mode == HW_ACC_MIG_VF_CTRL && func_len < PAGE_SIZE) + pci_warn(pdev, "migration not supported on 64KB pages with QM_HW_V3\n"); else - pci_warn(pdev, "migration support failed, continue with generic interface\n"); + ops = &hisi_acc_vfio_pci_migrn_ops; } hisi_acc_vdev = vfio_alloc_device(hisi_acc_vf_core_device, @@ -1691,6 +1697,8 @@ static int hisi_acc_vfio_pci_probe(struct pci_dev *pdev, const struct pci_device if (IS_ERR(hisi_acc_vdev)) return PTR_ERR(hisi_acc_vdev); + hisi_acc_vdev->drv_mode = drv_mode; + dev_set_drvdata(&pdev->dev, &hisi_acc_vdev->core_device); ret = vfio_pci_core_register_device(&hisi_acc_vdev->core_device); if (ret) -- 2.43.0