Greg Lindahl wrote:
>
> > I take it these
> > messages indicate the hang is around 2 seconds, which makes sense
> > given I send 16MB over 100Mb Enet,
>
> That depends on the detail. Are you running PVM in daemon mode or using
> direct routing? Even if you were sending 16 MB in a single blocking send, it
> should not hold the kernel lock while you're doing it. If so, it would be a
> generic kernel bug...
>
It would depend on exactly what those messages are trying to
say. I would expect pvmd to go into a hard read() loop of some
sort collecting bytes of the data until it gets the complete
chunk transferred. At least, that's how I write sockets code
when passing data. The reason being select() will return
readable on a single byte of data so it's basically no better
than just sitting on a read(). Nothing stopping one from using
select() but it doesn't buy you much unless you want to interleave
read()'ing multiple data streams. How the kernel locks work I
don't know but I assume there is a data structure somewhere that
is effected by the socket I/O that access to must be serialized
to keep things sane.
> strace is a very good utility (if it's functioning on alphas these days) to
> figure out what kinds of system calls you're making.
>
Haven't tried strace lately.
Wes
-- To unsubscribe: send e-mail to axp-list-request@redhat.com with 'unsubscribe' as the subject. Do not send it to axp-list@redhat.com
This archive was generated by hypermail version 2a22 on Sat Jul 1 05:31:30 2000 PDT
Send any problems or questions about this archive to webmaster@alphalinux.org.