On Mon, 15 Jul 2019 06:20:27 -0700
Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
> This looks good to me - if you and Tom are fine with it I'd like to
> fold it into his commit so that what I'll send to Linus is bisection
> clean.
No objections here.
>
> > Note: we still need sev_active() defined because of the reference
> > in fs/core/vmcore, but this one is likely to go away soon along
> > with the need for an s390 sev_active().
>
> Any chance we could not change the return value from the function
> at least in this patch/fold as that change seems unrelated to the
> dma functionality. If that is what you really wanted and only
> the dma code was in the way we can happily merge it as a separate
> patch, of couse.
>
AFAIU the story form fs/core/vmcore.c boils down to the same on s390. I
explained this in an email I've sent a moment ago (to Thiago). I expect
sev_active() on s390 to go away soon as it really does not make sense
for us (any more).
Thus yes, we can restore the pre- e67a5ed1f86f ("dma-direct: Force
unencrypted DMA under SME for certain DMA masks") sev_active() behavior
as well, even if we don't care about it. What I did conveys the semantic
better. Not changing the behavior of however sev_active() makes more
sense if the two are going to be squashed.
The corresponding diff looks like follows. Would you like me to send it
out as v2?
Regards,
Halil
----------------------------8<---------------------------------------
diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 5d8570ed6cab..a4ad2733eedf 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -189,6 +189,7 @@ config S390
select VIRT_CPU_ACCOUNTING
select ARCH_HAS_SCALED_CPUTIME
select HAVE_NMI
+ select ARCH_HAS_FORCE_DMA_UNENCRYPTED
select SWIOTLB
select GENERIC_ALLOCATOR
diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c
index f0bee6af3960..dfe47a22480a 100644
--- a/arch/s390/mm/init.c
+++ b/arch/s390/mm/init.c
@@ -155,12 +155,17 @@ int set_memory_decrypted(unsigned long addr, int numpages)
return 0;
}
-/* are we a protected virtualization guest? */
bool sev_active(void)
{
return is_prot_virt_guest();
}
+/* are we a protected virtualization guest? */
+bool force_dma_unencrypted(struct device *dev)
+{
+ return is_prot_virt_guest();
+}
+
/* protected virtualization */
static void pv_init(void)
{
