Date: Thu, 13 Feb 2014 16:05:14 -0500 From: Garrett Wollman <wollman@bimajority.org> To: John Baldwin <jhb@freebsd.org> Cc: FreeBSD Net <freebsd-net@freebsd.org> Subject: Re: Use of contiguous physical memory in cxgbe driver Message-ID: <21245.13194.145088.287817@hergotha.csail.mit.edu> In-Reply-To: <6154054.5G06XIekg6@ralph.baldwin.cx> References: <21216.22944.314697.179039@hergotha.csail.mit.edu> <201402121446.19278.jhb@freebsd.org> <CAJ-VmokDb4dkhzO6NnYrFKKV2g-bau8H47h2g=wOkvGcqYzRow@mail.gmail.com> <6154054.5G06XIekg6@ralph.baldwin.cx>
next in thread | previous in thread | raw e-mail | index | archive | help
<<On Thu, 13 Feb 2014 15:16:54 -0500, John Baldwin <jhb@freebsd.org> said: > It seeks to use the smallest size possible however. It is true that > we don't attempt to move a busy page elsewhere to free up memory > (e.g. if you had a 2MB free chunk with one busy 4k page in the > middle), but we can't really do that safely. Given the existence of > the direct map, we can't relocate a page and be sure that we have > also relocated all possible pointers to it. I doubt that this can be usefully auto-configured. As an administrator, I would be perfectly happy with a situation where I can tell the system to set aside some amount of memory at initialization (say, 16 superpages per (virtual) CPU) and get better performance. It's not a real imposition on a 128-GB server to set aside 1 GB of memory for big contiguous allocations if the applications will benefit from it. But in the absence of a guaranteed allocation, network drivers need to avoid requiring contiguous physical memory when they don't need it. -GAWollman
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?21245.13194.145088.287817>