Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 18 Dec 98 09:31:32 -0600
From:      "Richard Seaman, Jr." <lists@tar.com>
To:        "Julian Elischer" <julian@whistle.com>, "Mike Smith" <mike@smith.net.au>
Cc:        "current@freebsd.org" <current@FreeBSD.ORG>
Subject:   Re: Linux Threads patches available
Message-ID:  <199812181531.JAA10366@ns.tar.com>

next in thread | raw e-mail | index | archive | help
On Thu, 17 Dec 1998 19:57:30 -0800, Julian Elischer wrote:

>The only downside I see to this is the addition of 
>328 bytes to the size of the proc struct. 
>when those structs were in the U area, they were swapable.
>
>mind you, how many processes are swapped entirely on most systems?

When Brian Feldman posted the first version of his patches for
shared sigacts, someone (I'd have to check the archives to find
out who) suggested leaving the sigacts structure in the U area,
and only allocating the additional proc area structure in the event 
of an rforked process with the RFSIGSHARE option.  This would
involve then copying the values from the U area into the new
malloc'd structure and resetting the p_sigacts pointer in both
the parent and child.  Then when the refcnt of the proc area
structure gets back down to one, the values are copied back
out to the U area structure, the proc pointer is reset and
the area freed.

It involves a little more coding, but would avoid most of the 
increase in the proc structure for non-threaded processes (but
involves lugging around an unused U area structure per thread 
(ie. per process) for the RFSIGSHARE rforked process.  If
the extra proc size is a problem, this idea could be pursued.




To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199812181531.JAA10366>