On Wed, Mar 22, 2023 at 06:02:34PM +0100, David Hildenbrand wrote: > On 22.03.23 15:55, Lorenzo Stoakes wrote: > > Having previously laid the foundation for converting vread() to an iterator > > function, pull the trigger and do so. > > > > This patch attempts to provide minimal refactoring and to reflect the > > existing logic as best we can, for example we continue to zero portions of > > memory not read, as before. > > > > Overall, there should be no functional difference other than a performance > > improvement in /proc/kcore access to vmalloc regions. > > > > Now we have eliminated the need for a bounce buffer in read_kcore_iter(), > > we dispense with it, and try to write to user memory optimistically but > > with faults disabled via copy_page_to_iter_nofault(). We already have > > preemption disabled by holding a spin lock. > > > > If this fails, we fault in and retry a single time. This is a conservative > > approach intended to avoid spinning on vread_iter() if we repeatedly > > encouter issues reading from it. > > I have to ask again: Can you comment why that is ok? You might end up > signaling -EFAULT to user space simply because swapping/page > migration/whatever triggered at the wrong time. > > That could break existing user space or which important part am I missing? > Actually you're right, this is not ok. I was being mistakenly overcautious about spinning but in actual fact I don't think this would be an issue here. I will respin with a while loop so under no odd timing circumstance do we break userland. > -- > Thanks, > > David / dhildenb >
