In commit
1381a0c67671 ("s390/pci: Implement ioremap_wc/prot() with
MIO") we implemented both ioremap_wc() and ioremap_prot() however until
now we had not set HAVE_IOREMAP_PROT in Kconfig, do so now.
This also requires implementing pte_pgprot() as this is used in the
generic_access_phys() code enabled by CONFIG_HAVE_IOREMAP_PROT. As with
ioremap_wc() we need to take the MMIO Write Back bit index into account.
Moreover since the pgprot value returned from pte_pgprot() is to be used
for mappings into kernel address space we must make sure that it uses
appropriate kernel page table protection bits. In particular a pgprot
value originally coming from userspace could have the _PAGE_PROTECT
bit set to enable fault based dirty bit accounting which would then make
the mapping inaccessible when used in kernel address space.
Fixes: 1381a0c67671 ("s390/pci: Implement ioremap_wc/prot() with MIO")
Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
select HAVE_FUTEX_CMPXCHG if FUTEX
select HAVE_GCC_PLUGINS
select HAVE_GENERIC_VDSO
+ select HAVE_IOREMAP_PROT if PCI
select HAVE_IRQ_EXIT_ON_IRQ_STACK
select HAVE_KERNEL_BZIP2
select HAVE_KERNEL_GZIP
return pte_val(pte) & _PAGE_UNUSED;
}
+/*
+ * Extract the pgprot value from the given pte while at the same time making it
+ * usable for kernel address space mappings where fault driven dirty and
+ * young/old accounting is not supported, i.e _PAGE_PROTECT and _PAGE_INVALID
+ * must not be set.
+ */
+static inline pgprot_t pte_pgprot(pte_t pte)
+{
+ unsigned long pte_flags = pte_val(pte) & _PAGE_CHG_MASK;
+
+ if (pte_write(pte))
+ pte_flags |= pgprot_val(PAGE_KERNEL);
+ else
+ pte_flags |= pgprot_val(PAGE_KERNEL_RO);
+ pte_flags |= pte_val(pte) & mio_wb_bit_mask;
+
+ return __pgprot(pte_flags);
+}
+
/*
* pgd/pmd/pte modification functions
*/