Since it seems like a decidedly useful piece of functionality to allow a UMA zone's init/ctor function to cause zalloc failure, and it makes good sense to pass in the M_*WAIT* flags for their use, I have done so in a tree and found it really didn't seem to be a hard change to make. I think if we're going to do it, we need to do it before 5 is branched and not have these problems lurking. I've been running with this on my system with no ill effects so far, but it needs testing under load (i.e. from people that are able to truly provoke the panic: mb_ctor_pack(): can't deal with failure! Please try this if you are having that problem. That panic is possible for anyone with extreme mbuf/kmem load, and there are further panics in the case of users of the MAC subsystem when running out of memory in a similar fashion. <URL:http://green.homeunix.org/~green/uma_error_checking.patch> -- Brian Fundakowski Feldman \'[ FreeBSD ]''''''''''\ <> green_at_FreeBSD.org \ The Power to Serve! \ Opinions expressed are my own. \,,,,,,,,,,,,,,,,,,,,,,\Received on Fri Jul 02 2004 - 16:10:08 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:00 UTC