Re: 2.2.0-final, ppp 2.3.5 -> awful?

David Mansfield (david@cobite.com)
Tue, 26 Jan 1999 10:32:10 -0500 (EST)


Matthew Harrell <mharrell@std.saic.com> wrote:
>Frank McIngvale was overheard saying:
>: I was having the same problem. The PPP link would just stall after a
>while
>: and never come back. The link was up, and tcpdump showed a lot of ACK's
>: going out, but nothing coming in. On a whim, I turned off DMA on my IDE
>: hard drive, and the stalls have not come back (yet). But, I may just be
>: having a run of good luck.

>Those are pretty much the same symptoms I was having. Most of the time
>my link would come back up, though, but it took a while. I'll try out
>the dma setting - - how did you turn it off, hdparm?

The issue is that the IDE driver is taking too long to process a given
request and so the serial driver misses an interrupt and some (serial)
data is lost. Other contributing factors are the multiblock fetch count
(if set to high there is more data to transfer per interrupt I think) and
MOST IMPORTANTLY (for me and others) unmasking other interrupts while
procossing a request. All of these are settable using hdparm.

David

-- 
/==============================\
| David Mansfield              |
| david@cobite.com             |
\==============================/

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/