The current ls2/gram source trees do not wire the mandatory clock control / reset signals from the DDR PHY to the CRG.
(In reply to tpearson from comment #0) > The current ls2/gram source trees do not wire the mandatory clock control / > reset signals from the DDR PHY to the CRG. oink? how is anything working at all, then!
(In reply to Luke Kenneth Casson Leighton from comment #1) > > oink? how is anything working at all, then! Random chance, basically. Fix incoming that reduces the failure mode to bistable state, and the root cause is fascinating for the remaining problems....
Fixed in GIT hashes 7cb3e51d2d2c7e1d71fc9c991697e1270f60358b (gram) and 19ed0026e91b2dd351fbd2d692fb2c6f45b42622 (ls2). Basically, the reset system for GRAM right now is a complete mess. You can sorta get away with it as long as all the blocks come up on the correct SCLK phase, which happens more often than you might think just from random behavior.