Hi,
I noticed a strange bug while debugging flashrom on the DLP-USB1232H
board. Basically, flashrom uses the MPSSE SPI interface, sends 4 bytes
SPI command, then reads 256 bytes SPI response from a chip attached to
the FT2232H chip on that board.
The common pattern in the broken case (with very few deviations) is
(counting starts at 0)
byte 0-61 are read just fine, byte 62-63 are skipped,
byte 64-125 are read just fine, byte 126-127 are skipped,
byte 128-189 are read just fine, byte 190-191 are skipped,
byte 192-253 are read just fine, byte 254-255 are skipped,
the last 8 bytes in the buffer are filled with zeros (or unchanged).
Basically, only the first 62 bytes of every 64 byte block are read and
the next 64 byte block starts two bytes too early.
Only a part of the transfers is broken that way, most are OK. It is
interesting that the broken transfers tend to cluster around specific
addresses.
We can circumvent the issue by limiting chunk size to 32 bytes, but
that's not a real solution. None of the libftdi transfer functions
return any error, so I was expecting the transfers to be reliable.
Jeremy (who owns the hardware) is in CC and he'll supply details about
libftdi, libusb, kernel and hardware soon.
Uwe owns the same hardware, but he is not seeing any issues yet (well at
least not when keeping the machine load low).
Regards,
Carl-Daniel
--
Developer quote of the month:
"We are juggling too many chainsaws and flaming arrows and tigers."
--
libftdi - see http://www.intra2net.com/en/developer/libftdi for details.
To unsubscribe send a mail to libftdi+unsubscribe@xxxxxxxxxxxxxxxxxxxxxxx
|
