On Wed, Mar 20, 2019 at 08:55:17AM -0600, William Kucharski wrote: > > > > On Mar 19, 2019, at 10:33 PM, Jerome Glisse <jglisse@xxxxxxxxxx> wrote: > > > > So i believe best we could do is send a SIGBUS to the process that has > > GUPed a range of a file that is being truncated this would match what > > we do for CPU acces. There is no reason access through GUP should be > > handled any differently. > > This should be done lazily, as there's no need to send the SIGBUS unless > the GUPed page is actually accessed post-truncate. Issue is that unlike CPU access we might not be able to detect device access and thus it is not something we can do lazily for everyone. Cheers, Jérôme
