From: Denis V. <vd...@il...> - 2005-10-06 14:01:18
|
On Thursday 06 October 2005 16:25, Jason Grant wrote: > On Thu, 2005-10-06 at 13:50 +0300, Denis Vlasenko wrote: > > Well, we have very scarce info on acx protocol, > > and what we have is contradictory in places. > > OK then, I'll give it a go tomorrow & report back. > > > What leak? We have statically allocated tx buffers. > > Not a memory leak, I meant a possible 'tx descriptor' leak, and yes, it > can only occur to the point where the ring buffer is saturated, which is > what I'm seeing. When the driver is working OK, the ACXDONE bit is > being reset, however the fault that I'm encountering involves this bit > not being reset. Anyway, the idea about such a leak is only speculation > - another possibility is general queue overflow as mentioned earlier. BTW, consider lowering this: #define TX_CNT 32 to 16 or even 8 -- vda |