On Mo, 2019-05-20 at 10:16 -0400, Alan Stern wrote: > On Mon, 20 May 2019, Christoph Hellwig wrote: > > > GFP_KERNEL if you can block, GFP_ATOMIC if you can't for a good reason, > > that is the allocation is from irq context or under a spinlock. If you > > think you have a case where you think you don't want to block, but it > > is not because of the above reasons we need to have a chat about the > > details. > > What if the allocation requires the kernel to swap some old pages out > to the backing store, but the backing store is on the device that the > driver is managing? The swap can't take place until the current I/O > operation is complete (assuming the driver can handle only one I/O > operation at a time), and the current operation can't complete until > the old pages are swapped out. Result: deadlock. > > Isn't that the whole reason for using GFP_NOIO in the first place? Hi, lookig at this it seems to me that we are in danger of a deadlock - during reset - devices cannot do IO while being reset covered by the USB layer in usb_reset_device - resume & restore - devices cannot do IO while suspended covered by driver core in rpm_callback - disconnect - a disconnected device cannot do IO is this a theoretical case or should I do something to the driver core? How about changing configurations on USB? Regards Oliver
