Re: LOR from yesterday

From: Jeffrey Hsu <hsu_at_FreeBSD.org>
Date: Thu, 01 May 2003 04:34:08 -0700
  soreceive(c66e5400,dee6eb1c,dee6eb28,dee6eb20,0) at soreceive+0x819
  nfsrv_rcv(c66e5400,c6c0fe80,4) at nfsrv_rcv+0x75
  sowakeup(c66e5400,c66e544c) at sowakeup+0x7f
  tcp_input(c1f75600,14) at tcp_input+0x11f0

Yeah, that one's been there for almost a year now.  A long time ago,
I made a fix for it which you can find at 
	http://people.freebsd.org/~hsu/hammer.diff
It takes a big hammer to the problem and defers all the upcalls
until the lock is released.  However, I don't like it much and was
hoping for a unified solution which can take care of all socket upcalls,
not just NFS.  And also, once we turn Giant off around the networking
stack, it won't be safe to run NFS anymore, until that subsystem gets
locked up.  So, all in all, it just seems better to defer this issue
until then.

							Jeffrey
Received on Thu May 01 2003 - 02:24:19 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:05 UTC