On 19.10.22 14:31, David Hildenbrand wrote:
On 27.09.22 13:01, David Hildenbrand wrote:
Let's trigger a R/O longterm pin on three cases of R/O mapped anonymous pages:
- exclusive (never shared)
- shared (child still alive)
- previously shared (child no longer alive)
... and make sure that the pin is reliable: whatever we write via the page tables has to be observable via the pin.
Signed-off-by: David Hildenbrand david@redhat.com
The following on top should silence the warning on 32bit:
https://lore.kernel.org/all/202210182101.XOyPZEQV-lkp@intel.com/T/#u
diff --git a/mm/gup_test.c b/mm/gup_test.c index 2bbfacd31f5a..57c343f2dc51 100644 --- a/mm/gup_test.c +++ b/mm/gup_test.c @@ -299,7 +299,8 @@ static inline int pin_longterm_test_read(unsigned long arg) for (i = 0; i < pin_longterm_test_nr_pages; i++) { void *addr = page_to_virt(pin_longterm_test_pages[i]);
if (copy_to_user((void __user *)user_addr, addr, PAGE_SIZE))
if (copy_to_user((void __user *)(unsigned long)user_addr, addr,
PAGE_SIZE)) return -EFAULT; user_addr += PAGE_SIZE; }
Ehm, wrong patch, it actually belongs into #6/7
https://lkml.kernel.org/r/20220927110120.106906-7-david@redhat.com
@Andrew, if your confused, I can just resend the whole thing.