Messages in this thread | | | Date | Sat, 5 Oct 2013 08:51:48 -0700 | From | Christoph Hellwig <> | Subject | Re: lustre: why does cfs_get_random_bytes() exist? |
| |
On Sat, Oct 05, 2013 at 10:21:21AM -0400, Theodore Ts'o wrote: > add_device_randomness() is called from __dev_open() and > dev_set_mac_address() in net/core/dev.c. This is above the ethernet > and infiniband level. So as long as it looks like a Linux network > device, and they are setting the hardware media access address in the > standard place (dev->dev_addr), it should work fine. > > If they don't then they should fix their drivers to call > add_device_randomness(); the answer shouldn't be to make every single > users of the Linux random number generation infrastructure work around > the problem at the subsystem or file system level!
Besides these points a filesystem isn't the place to work around entropy pool issues on some obscure platform. Even if they can't fix it this way for some reason the workaround still should be in the interaction of the arch code and the Linux prng infrastructure and not in Lustre.
| |