Re: port-i386/33291

[T <bobs%thelibertytree.org@localhost>]

The following reply was made to PR port-i386/33291; it has been noted by GNATS.

From: T <bobs%thelibertytree.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: "abs%absd.org@localhost >> David Brownlee" <abs%absd.org@localhost>
Subject: Re: port-i386/33291
Date: Sat, 16 Nov 2019 02:31:56 -0700

 Hello, I too was having this problem with the pdcsata driver for a 
 PDC40719 FastTrak TX4. I am using a different port (NetBSD/prep) but 
 would have similar errors to:
 
 pdcsata0:1:0: lost interrupt
  Â Â Â Â Â Â Â  type: ata tc_bcount: 16384 tc_skip:01
 
 the difference being the 'tc_bcount' number and zero 'tc_skip'. 
 Conducting certain tasks with high drive IO would eventually result in a 
 "stuck" terminal  (DDB could be entered using the key sequence, caps/num 
 lock LED would toggle, machine could be pinged). I found that adding the 
 'buffer queue strategy' option BUFQ_READPRIO to the kernel config 
 resolved these problems and have not seen them since using it. The other 
 option BUFQ_PRIOSCAN may work, but it is unknown to me at this point, 
 because it will cause a kernel panic if enabled for this specific 
 system. It should also be noted that many ports have enabled 
 BUFQ_PRIOSCAN in their GENERIC configs, including i386, so this problem 
 may not resurface, unless someone specifically uses the 'disksort' 
 strategy in combination with this driver. This is an old ticket, but 
 figured this information would help others that may run into this problem.
 
 -Tim