On Wed, Dec 11, 2019 at 4:05 AM Lakshmi Ramasubramanian <nramas@xxxxxxxxxxxxxxxxxxx> wrote: > > existing in-kernel measurement that is kept secure. In other words, > > each exported list would have a measurement of the earlier generation > > list and we would build a simple list chain. > > Do we need to keep multiple on-disk lists? Can the measurement entries > be written to one on-disk file - say, when the current in-memory buffer > reaches a certain threshold? Ok, one way is to append indeed. Maybe this would be simpler since every new boot could just overwrite the old file. > A remote attestation of the system would then read the on-disk file and > the current in-memory buffer to perform the validation. > > I am assuming one of the reasons for keeping the measurement list in > memory is for better performance. If buffered file I/O is supported in > Linux, can that be leveraged for improved file I/O performance? AFAIK Linux only does buffered io (with LRU) unless sync is being explicitly asked. Exporting the file would be a rare event, so it will not hopefully be a performance sensitive operation by any means. -- Janne
