I'm trying to develop an RS-232 protocol sniffer that lets a DB9 cable 
pass through my board and captures all the communication in both 
directions.  I want to log all this data to an onboard USB flash drive, 
and I'm having trouble coming to terms with something that happens in 
the FAT implementation.  Essentially, around every 32K characters sent 
(I'm buffering between 400 and 512 bytes per write), the library code 
blocks for up to 1.2 seconds.  I've tracked it down to a call to 
fsfwrite and then to flushdata, then to USBHostMSDSCSISectorWrite which 
has this blocking code :

while (!USBHostMSDTransferIsComplete( deviceAddress, &errorCode, 
&byteCount ))
{
    USBTasks();
}

So apparently, the library code just blocks while waiting for the mass 
storage device to complete the data transfer.  Does anyone have a 
suggestion about how to handle this?  I'm buffering the incoming 
characters in interrupts so they don't get lost, but I was hoping to log 
them to the drive with timestamps, and buffering the timestamp along 
with every incoming character for 1 second gets pretty cumbersome in 
terms of ram usage (especially at high monitored baud rates).  I know I 
can work around it for low data rates (I've been trying to think of a 
relative timestamp that could just be a byte or two per character 
instead of the whole date time), but I'm wondering if there are better 
approaches than the Microchip library for this, and also whether anyone 
has experienced this kind of time delay when logging to flash drives.

Thanks,

Tony



-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist