Michael S. Tsirkin
2020-Oct-01 12:46 UTC
[PATCH v2 0/2] vhost: Skip access checks on GIOVAs
On Tue, Sep 29, 2020 at 06:30:20PM +0200, Greg Kurz wrote:> This series addresses some misuse around vring addresses provided by > userspace when using an IOTLB device. The misuse cause failures of > the VHOST_SET_VRING_ADDR ioctl on POWER, which in turn causes QEMU > to crash at migration time. > > While digging some more I realized that log_access_ok() can also be > passed a GIOVA (vq->log_addr) even though log_used() will never log > anything at that address. I could observe addresses beyond the end > of the log bitmap being passed to access_ok(), but it didn't have any > impact because the addresses were still acceptable from an access_ok() > standpoint. Adding a second patch to fix that anyway. > > Note that I've also posted a patch for QEMU so that it skips the used > structure GIOVA when allocating the log bitmap. Otherwise QEMU fails to > allocate it because POWER puts GIOVAs very high in the address space (ie. > over 0x800000000000000ULL). > > https://patchwork.ozlabs.org/project/qemu-devel/patch/160105498386.68108.2145229309875282336.stgit at bahia.lan/I queued this. Jason, can you ack please?> v2: > - patch 1: move the (vq->ioltb) check from vhost_vq_access_ok() to > vq_access_ok() as suggested by MST > - patch 2: new patch > > --- > > Greg Kurz (2): > vhost: Don't call access_ok() when using IOTLB > vhost: Don't call log_access_ok() when using IOTLB > > > drivers/vhost/vhost.c | 32 ++++++++++++++++++++++++-------- > 1 file changed, 24 insertions(+), 8 deletions(-) > > -- > Greg