Date: Mon, 20 Jul 2015 10:47:58 -0700 From: John Baldwin <jhb@freebsd.org> To: Alan Cox <alc@rice.edu> Cc: Adrian Chadd <adrian@freebsd.org>, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r285387 - in head: lib/libc/sys share/man/man4 sys/conf sys/kern sys/sys sys/vm usr.bin usr.bin/numactl Message-ID: <14982298.P5Y2Xh6Asn@ralph.baldwin.cx> In-Reply-To: <55A1445F.50901@rice.edu> References: <201507111521.t6BFLcrv039934@repo.freebsd.org> <55A1445F.50901@rice.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Saturday, July 11, 2015 11:29:19 AM Alan Cox wrote: > On 07/11/2015 10:21, Adrian Chadd wrote: > > * The VM doesn't handle unbalanced domains very well, and if you have an overly > > unbalanced memory setup whilst under high memory pressure, VM page allocation > > may fail leading to a kernel panic. This was a problem in the past, but it's > > much more easily triggered now with these tools. > > > > > For the record, no, it doesn't panic. Both the first-touch scheme in > 9.x and the round-robin scheme in 10.x fall back to allocating from a > different domain until some page is found. I got a panic (don't recall exactly which) with the 9.x version once (albeit with that version backported to 8.x) where vm_page_alloc() returned NULL when a caller did not expect it based on the global paging targets. Unfortunately I no longer have access to the core (or any notes I might have had from debugging it). I don't recall why it didn't fall back to using a page from another domain (especially since the 9.x version just prefers local, doesn't require local). Note that I only saw this once across hundreds of machines running the 9.x version in production for 5 years or so. -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?14982298.P5Y2Xh6Asn>