From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alex Williamson Subject: Re: [RFC PATCH 3/6] VFIO: unregister IOMMU notifier on error recovery path Date: Tue, 13 Nov 2012 11:15:51 -0700 Message-ID: <1352830551.2233.18.camel@bling.home> References: <1352555839-18961-1-git-send-email-jiang.liu@huawei.com> <1352555839-18961-3-git-send-email-jiang.liu@huawei.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: Greg Kroah-Hartman , Jiang Liu , Joerg Roedel , Yinghai Lu , kvm@vger.kernel.org, linux-kernel@vger.kernel.org, Hanjun Guo To: Jiang Liu Return-path: In-Reply-To: <1352555839-18961-3-git-send-email-jiang.liu@huawei.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On Sat, 2012-11-10 at 21:57 +0800, Jiang Liu wrote: > From: Jiang Liu > > On error recovery path in function vfio_create_group(), it should > unregister the IOMMU notifier for the new VFIO group. Otherwise it may > cause invalid memory access later when handling bus notifications. > > Signed-off-by: Jiang Liu > --- > drivers/vfio/vfio.c | 31 +++++++++++++++---------------- > 1 file changed, 15 insertions(+), 16 deletions(-) This patch and patch 6/6 looks like good vfio fixes regardless of how we tackle the driver binding problem. Please submit them separately. Thanks for the patches, I look forward to a solution here. Thanks, Alex > > diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c > index 17830c9..3359ec2 100644 > --- a/drivers/vfio/vfio.c > +++ b/drivers/vfio/vfio.c > @@ -191,6 +191,17 @@ static void vfio_container_put(struct vfio_container *container) > kref_put(&container->kref, vfio_container_release); > } > > +static void vfio_group_unlock_and_free(struct vfio_group *group) > +{ > + mutex_unlock(&vfio.group_lock); > + /* > + * Unregister outside of lock. A spurious callback is harmless now > + * that the group is no longer in vfio.group_list. > + */ > + iommu_group_unregister_notifier(group->iommu_group, &group->nb); > + kfree(group); > +} > + > /** > * Group objects - create, release, get, put, search > */ > @@ -229,8 +240,7 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group) > > minor = vfio_alloc_group_minor(group); > if (minor < 0) { > - mutex_unlock(&vfio.group_lock); > - kfree(group); > + vfio_group_unlock_and_free(group); > return ERR_PTR(minor); > } > > @@ -239,8 +249,7 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group) > if (tmp->iommu_group == iommu_group) { > vfio_group_get(tmp); > vfio_free_group_minor(minor); > - mutex_unlock(&vfio.group_lock); > - kfree(group); > + vfio_group_unlock_and_free(group); > return tmp; > } > } > @@ -249,8 +258,7 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group) > group, "%d", iommu_group_id(iommu_group)); > if (IS_ERR(dev)) { > vfio_free_group_minor(minor); > - mutex_unlock(&vfio.group_lock); > - kfree(group); > + vfio_group_unlock_and_free(group); > return (struct vfio_group *)dev; /* ERR_PTR */ > } > > @@ -274,16 +282,7 @@ static void vfio_group_release(struct kref *kref) > device_destroy(vfio.class, MKDEV(MAJOR(vfio.devt), group->minor)); > list_del(&group->vfio_next); > vfio_free_group_minor(group->minor); > - > - mutex_unlock(&vfio.group_lock); > - > - /* > - * Unregister outside of lock. A spurious callback is harmless now > - * that the group is no longer in vfio.group_list. > - */ > - iommu_group_unregister_notifier(group->iommu_group, &group->nb); > - > - kfree(group); > + vfio_group_unlock_and_free(group); > } > > static void vfio_group_put(struct vfio_group *group)