Date: Thu, 15 Aug 1996 06:51:14 +1000 From: Bruce Evans <bde@zeta.org.au> To: mrm@Mole.ORG, nate@mt.sri.com Cc: hackers@FreeBSD.ORG Subject: Re: sio issues (silo overflows on a pentium, locked in ttywait, etc...) Message-ID: <199608142051.GAA03505@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>I'm still having problems with > sio1: 182 more interrupt-level buffer overflows (total 883) >Each one causes an Ierr visible with netstat -ai. The system >recovers and keeps going, so I can't complain (too much :-). Interrupt-level buffer overflows are caused by something masking softtty interrupts for too long. The buffer has space for about 196 extra characters at your speed of 57600 bps and overrunning it by 182 characters at 57600 bps means that something has disabled softty interrupts for more than 65 ms. softty interrupts are disabled mainly when hard tty interrupts are disabled - i.e., for critical clock, vm, network and tty processing (net_imask includes tty_imask if splip or ppp is configured). >The system is a 2.1R 16MB 486DX2/50 with 16550 _clones_ on a cheap IDE, >a 1542B, and I've added >options "TTYHOG=4096" # to fix buffer overflows in sio.c >options "RS_IBUF_SIZE=1024" # to fix buffer overflows in sio.c >... >This machine is still running FTL 14. I don't get any silo overflows. >Adding those options really cut down on the interrupt-buffer overflows. >Looks like about 10-15% of what I was getting before. Still there, though. If you still get them with an RS_IBUFSIZE of 1024, then something must be disabling softtty interrupts for more than 199 ms :-(. Perhaps there is a missing splx(). Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199608142051.GAA03505>