Date: Mon, 18 Jan 1999 20:30:22 -0500 (EST) From: "John S. Dyson" <root@dyson.iquest.net> To: dillon@apollo.backplane.com (Matthew Dillon) Cc: dg@root.com, jkh@FreeBSD.ORG, hackers@FreeBSD.ORG Subject: Re: Found problem w/ Paging performance over NFS Message-ID: <199901190130.UAA27297@dyson.iquest.net> In-Reply-To: <199901190128.RAA84794@apollo.backplane.com> from Matthew Dillon at "Jan 18, 99 05:28:35 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
> : > :PG_BUSY is quite severe, and not always needed. > : > :John > > Ok. I just figured out the interaction -- in fact, we *can't* test m->busy > in this section of code because of the following sequence: > > vm_fault -> vnode_pager_getpages -> ffs_getpages -> ffs_read -> > cluster_read -> getblk -> allocbuf ( lockup ) > > ffs_getpages converts the PG_BUSY into a vm_page_t->busy before getting into > the cluster stuff, and thus allocbuf() deep down cannot sleep on > vm_page_t->busy. > > Ick. I'm going to add a comment explaining that case. > Yep, it is *very* important to understand the interactions, or either clustering won't work on alot of filesystems or worse things will happen. John To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901190130.UAA27297>