Date: Thu, 21 Jan 1999 21:49:41 -0800 (PST) From: Matt Dillon <dillon@FreeBSD.ORG> To: cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: cvs commit: src/sys/vm vm_fault.c Message-ID: <199901220549.VAA29094@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
dillon 1999/01/21 21:49:40 PST
Modified files: (Branch: RELENG_3)
sys/vm vm_fault.c
Log:
Backport from -4.x. We are supposed to wait while the page is busy before
messing with it, but the VM code was special-casing a m->busy page that
was still valid ( i.e. usually a pageout ) and allowing it through. The
problem is that whatever m->busy'd the page is probably still depending on
it not being messed with. Specifically, I am thinking of the
vm_pager_get_pages() code where the OBJT_VNODE pager converts PG_BUSY
to m->busy for the read operation. While this code may not be
specifically broken for the case, it does initially set PG_BUSY and assume
that it has the entire page to itself... but OBJT_VNODE converts it to
m->busy. It is just too dangerous to allow the special case optimization
through.
Removing the special optimization should have no discernable effect since
pageout operates mark resident pmap pte's READ-ONLY rather then unmapping
them, and so no fault normally occurs.
Revision Changes Path
1.93.2.1 +24 -9 src/sys/vm/vm_fault.c
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901220549.VAA29094>