Date: Sat, 31 May 1997 04:58:47 +1000 From: Bruce Evans <bde@zeta.org.au> To: bmcgover@cisco.com, luiz@nlink.com.br Cc: hackers@FreeBSD.ORG Subject: Re: Cyclades card problems... Message-ID: <199705301858.EAA17643@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>The problem is directly collated between interrupt latency and the number >of silo overflows. Yes. >The "problem" is that the Cyclom-Y cards only have DTR flow control >available to them. Note that this means no RTS/CTS flow control that most >applications use for hardware flow control. No. At least most of the ISA Cyclom-Y cards have RTS and CTS connected, and the driver supports this: Cyclom-8Ys: RJ-12 connector with CTS but no RTS. Also no RI (ring indicator). Only CTS output flow control works. Cyclom-8Yo,8Yb,16Y: DB-25 connector with everything except RI. RTS/CTS flow control works in both directions. The UARTs on the Cyclom-Y cards only support DTR flow control, and neither the cards nor the driver fudges this by pretending that DTR is actually RTS. Not much is lost, since the driver can't afford to depend on working RTS flow control anyway. E.g., the device might be connected to something that doesn't support it, or it might be connected to something that can't stop the flow fast enough. Any 16550-based device may have the latter problem (since the transmitter fifo size of 16 is larger than the receiver fifo size of 12). Anyway, flow control should be avoided if possible since it reduces throughput. >The PCI card, using a low priority interrupt handler, is the primary cause >for problems on the PCI version (the Ys have, if I remember correctly, as >this conversation was back in December, only 12 bytes of FIFO space).The >system just doesn't get to the card soon enough to avoid a silo overflow. > >In the ISA case, the problem is lessened, The problem is lessened by a large factor in the ISA case (maximum interrupt latency is typically 100 times smaller). This will be fixed when the PCI interrupt dispatcher supports fast interrupt handlers. >but an overburdened box will have >the same issues (and I've caused this in a pentium pro 200 running 16 ports >at 38400 at full saturation). This shouldn't happen. The driver is tuned to support 8 ports at 115200 at almost full saturation on a 486-33. (70% overhead, but not quite full saturation since the UARTs on Cyclom cards can't transmit at a full 115200). A P6/200 should have < 50% overhead for 16 ports at only 38400 saturated. Overhead is more of a problem than latency for the ISA driver. Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199705301858.EAA17643>