RE: [2.6.10-bk8] [SERIAL] dropping chars when > 512

["Stuart MacDonald" <stuartm@connecttech.com>]

From: Tim_T_Murphy@Dell.com
> examining the corrupt message contents, a group of bytes is 
> missing exactly at offset 512 in the message buffer.  
> Additional message bytes follow the gap, but the receiver's 
> checksum doesn't match the sender's, so the whole message is 
> dropped by the app.
> 
> I enabled debugging in 8250.c and serial_core.c.  Booting the 
> 2.6 kernel, the log shows:
> 
[snip]
> Jan  5 12:08:16 racrhel4 last message repeated 6 times
> Jan  5 12:08:16 racrhel4 kernel: 
> serial8250_interrupt(169)...status = 61...end.
> Jan  5 12:08:16 racrhel4 kernel: 
> serial8250_interrupt(169)...status = 60...THRE...end.
> Jan  5 12:08:16 racrhel4 last message repeated 3 times
> Jan  5 12:08:17 racrhel4 kernel: 
> serial8250_interrupt(169)...status = 61...end.
> Jan  5 12:08:17 racrhel4 kernel: 
> serial8250_interrupt(169)...status = 61...end.
> Jan  5 12:08:17 racrhel4 kernel: 
> serial8250_interrupt(169)...status = 61...THRE...end.
> Jan  5 12:08:17 racrhel4 kernel: 
> serial8250_interrupt(169)...status = 1...THRE...end.
> Jan  5 12:08:17 racrhel4 kernel: 

Unless this is a typo, I think you'll find that status = 1 means the
FIFOs have been turned off. Which would flush any data in the FIFOs.
Which would explain the missing data.

..Stu