vfio/nvlink: Remove exec permission to avoid SELinux AVCs
PATCH 1/1 vfio/nvlink: Remove exec permission to avoid SELinux AVCs
If SELinux is setup without 'execmem' permission for qemu, all mmap with (PROT_WRITE | PROT_EXEC) will fail and print a warning in SELinux log.
If "nvlink2-mr" memory allocation fails (fist diff), it will cause guest NUMA nodes to not be correctly configured (V100 memory will not be visible for guest, nor its NUMA nodes).
Not having 'execmem' permission is intesting for virtual machines to avoid buffer-overflow based attacks, and it's adopted in distros like RHEL.
So, removing the PROT_EXEC flag seems the right thing to do.
Browsing some other code that mmaps memory for usage with memory_region_init_ram_device_ptr, I could notice it's usual to not have PROT_EXEC (only PROT_READ | PROT_WRITE), so it should be no problem around this.
--- hw/vfio/pci-quirks.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
hw/vfio/pci-quirks.c +2 -2
@@ -1620,7 +1620,7 @@ int vfio_pci_nvidia_v100_ram_init(VFIOPCIDevice *vdev, Error **errp) } cap = (void *) hdr; - p =mmap (NULL , nv2reg->size,PROT_READ |PROT_WRITE |PROT_EXEC , + p =mmap (NULL , nv2reg->size,PROT_READ |PROT_WRITE ,MAP_SHARED , vdev->vbasedev.fd, nv2reg->offset);if (p ==MAP_FAILED ) { ret = -errno;
@@ -1680,7 +1680,7 @@ int vfio_pci_nvlink2_init(VFIOPCIDevice *vdev, Error **errp)/* Some NVLink bridges may not have assigned ATSD */ if (atsdreg->size) { - p =mmap (NULL , atsdreg->size,PROT_READ |PROT_WRITE |PROT_EXEC , + p =mmap (NULL , atsdreg->size,PROT_READ |PROT_WRITE ,MAP_SHARED , vdev->vbasedev.fd, atsdreg->offset);if (p ==MAP_FAILED ) { ret = -errno;
Signature · 1 line
2.25.4
Quoted 30 lines · Leonardo Bras
If SELinux is setup without 'execmem' permission for qemu, all mmap with (
PROT_WRITE|PROT_EXEC) will fail and print a warning in SELinux log.If "nvlink2-mr" memory allocation fails (fist diff), it will cause guest NUMA nodes to not be correctly configured (V100 memory will not be visible for guest, nor its NUMA nodes).
Not having 'execmem' permission is intesting for virtual machines to avoid buffer-overflow based attacks, and it's adopted in distros like RHEL.
So, removing the
PROT_EXECflag seems the right thing to do.Browsing some other code that mmaps memory for usage with
memory_region_init_ram_device_ptr, I could notice it's usual to not havePROT_EXEC(onlyPROT_READ|PROT_WRITE), so it should be no problem around this.Signed-off-by Leonardo Bras <[email protected]>--- hw/vfio/pci-quirks.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)hw/vfio/pci-quirks.c +0
-0
hw/vfio/pci-quirks.c int vfio_pci_nvidia_v100_ram_init(VFIOPCIDevice *vdev, Error **errp) · line 1623
Quoted 5 lines
hw/vfio/pci-quirks.c +1
-1@@ -1620,7 +1620,7 @@ int vfio_pci_nvidia_v100_ram_init(VFIOPCIDevice *vdev, Error **errp) } cap = (void *) hdr; - p = mmap(NULL, nv2reg->size, PROT_READ | PROT_WRITE | PROT_EXEC, + p = mmap(NULL, nv2reg->size, PROT_READ | PROT_WRITE,
Unlike other users of memory_region_init_ram_device_ptr(), this memory is more like a normal RAM, it is cache coherent and you can execute code from it. I am unaware if this happens in practice but it easily could.
Having said that, I can see QEMU is not setting PROT_EXEC even for RAM but KVM sets it anyway in a PTE in the partition scope tree (== guest physical-to-host physical translation table) at [1] which is fine, I just checked, with the change, the exec bit is still in the partition tree.
hw/vfio/pci-quirks.c int vfio_pci_nvlink2_init(VFIOPCIDevice *vdev, Error **errp) · line 1683
Quoted 9 lines
hw/vfio/pci-quirks.c +1
-1MAP_SHARED, vdev->vbasedev.fd, nv2reg->offset);if (p ==MAP_FAILED ) { ret = -errno;@@ -1680,7 +1680,7 @@ int vfio_pci_nvlink2_init(VFIOPCIDevice *vdev, Error **errp)/* Some NVLink bridges may not have assigned ATSD */ if (atsdreg->size) { - p = mmap(NULL, atsdreg->size, PROT_READ | PROT_WRITE | PROT_EXEC, + p = mmap(NULL, atsdreg->size, PROT_READ | PROT_WRITE,
This chunk is ok as ATSD is a set of registers.
Quoted 4 lines
hw/vfio/pci-quirks.c +0
-0MAP_SHARED, vdev->vbasedev.fd, atsdreg->offset);if (p ==MAP_FAILED ) { ret = -errno;
Signature · 1 line
Alexey