s390/pci: convert high_memory to physical address

We use high_memory as a measure for amount of memory available in
determining the required minimum size of our IOVA space with the
assumption that one rarely maps more than the available memory for DMA.
In special cases like mapping significant amounts of memory more than
once this can still be tuned with the s390_iommu_apterture kernel
parameter. In this use case high_memory is treated as a physical
address. As high_memory is a virtual address however this means we need
to convert it using virt_to_phys() before use

Note that at the moment physical and virtual addresses are identical so
this mismatch does not currently cause trouble.

Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
This commit is contained in:
Niklas Schnelle 2022-09-07 11:38:30 +02:00 committed by Vasily Gorbik
parent 5078775531
commit 2187582c36

View File

@ -666,7 +666,7 @@ static int __init dma_alloc_cpu_table_caches(void)
int __init zpci_dma_init(void)
{
s390_iommu_aperture = (u64)high_memory;
s390_iommu_aperture = (u64)virt_to_phys(high_memory);
if (!s390_iommu_aperture_factor)
s390_iommu_aperture = ULONG_MAX;
else