Do not emit an I/O page fault on amdvi_page_walk() when a valid mapping
is not found. The current role of amdvi_page_walk() is to be a helper
for the translate() method and ultimately the IOMMU replay()
functionality. These operations might be executed while the guest is
running, but do not necessarily correspond 1:1 with guest-initiated page
walks.
One example of such scenario is when VFIO code calls
memory_region_iommu_replay() (which ends up calling amdvi_walk_page())
to sync the dirty bitmap, or after registering a new notifier. The guest
would get IO_PAGE_FAULT events for all the regions where a mapping
doesn't currently exist, which is not correct.
Note that after this change there are no users of amdvi_page_fault(),
but since the IO page fault handling will be addressed in upcoming work,
I am choosing to mark it as unused rather than deleting it.
Signed-off-by: Alejandro Jimenez <alejandro.j.jime...@oracle.com>
---
hw/i386/amd_iommu.c | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/hw/i386/amd_iommu.c b/hw/i386/amd_iommu.c
index dc29a52bd845..ac7000dbafc7 100644
--- a/hw/i386/amd_iommu.c
+++ b/hw/i386/amd_iommu.c
@@ -263,8 +263,8 @@ static void amdvi_encode_event(uint64_t *evt, uint16_t
devid, uint64_t addr,
*
* @addr: virtual address in translation request
*/
-static void amdvi_page_fault(AMDVIState *s, uint16_t devid,
- hwaddr addr, uint16_t info)
+static void __attribute__((unused))
+amdvi_page_fault(AMDVIState *s, uint16_t devid, hwaddr addr, uint16_t info)
{
uint64_t evt[2];
@@ -1254,9 +1254,6 @@ static void amdvi_page_walk(AMDVIAddressSpace *as,
uint64_t *dte,
*/
if ((pte == (uint64_t)-1) || (pte == (uint64_t)-2) ||
!IOMMU_PTE_PRESENT(pte) || perms != (perms & amdvi_get_perms(pte))) {
-
- amdvi_page_fault(as->iommu_state, as->devfn, addr, perms);
- trace_amdvi_page_fault(addr);
return;
}
--
2.43.5
