Hi Chris, > > > > Amend the problem by keeping the mock file open throughout the > > > > duration of this test and calling fput() from the test itself. > > > > > > That is not the problem; that is papering over whatever the problem > > > really is. > > Would you mind sharing a bit more? At this point we have a > > problem in our code that we need to address and rejecting the > > proposed change without any suggestions gives me nothing to work > > with. > > The selftest leaks the vm_area_struct (and through that the object) on > any of the early failure paths. That seems unrelated to the commit > messages and the conclusions drawn from that false premise. > > Note in the vma_lookup() we could also do a check on > (addr + obj->base.size - 1) to make sure the entire range we expect to > have mapped is mapped to our target object. > Oh, I see! Thanks for the pointers, I did not consider that situation, but now that you mention it, it sounds spot on. I'll start working on v4 then.
Best Regards, Krzysztof
