>   Our question is: Is there an alternative path for "recvfrom" which is
 >   more simple than the current sotpi_recvmsg imeplementation, and will
 >   not lock during copying data from kernel to user space ?

I'm not familiar with the sotpi_recvmsg() codepaths, but a quick look
suggests that it would require some reorganization to avoid being under
SOREADLOCKED when doing the copyout, since that's done by kstrgetmsg().

Have you compared the performance against having a single thread do the
reads on the socket and then farming out the work to worker threads in
your application?

Hopefully someone with more familiarity in this area will have some
other ideas for you to try out.

--
meem
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to