On Wed, Dec 07, 2005, Andy Burns wrote: > On a similar subject the DVB API says that UCB will be an ever > increasing 32bit count of bad blocks that eventually wraps back past > zero, however in practice on my DVB-200T it seems to be a count of > errors between status reads, a quick look at the source of the TDA1004x > driver seems to confirm that the count is reset after being read, is the > spec wrong or the implementation? Well, all drivers are wrong. The spec is right, because with the actual implementation you get wrong numbers if more than one process opens the frontend read-only because the call to FE_READ_UNCORRECTED_BLOCKS resets the counter. (e.g. run two femon in parallel) For me this doesn't have enough practical relevance to fix it, though. Johannes
