[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZGZ5f00iemeG3GpC@moria.home.lan>
Date: Thu, 18 May 2023 15:16:15 -0400
From: Kent Overstreet <kent.overstreet@...ux.dev>
To: Song Liu <song@...nel.org>
Cc: Mike Rapoport <rppt@...nel.org>, linux-mm@...ck.org,
Andrew Morton <akpm@...ux-foundation.org>,
Dave Hansen <dave.hansen@...ux.intel.com>,
Peter Zijlstra <peterz@...radead.org>,
Rick Edgecombe <rick.p.edgecombe@...el.com>,
Thomas Gleixner <tglx@...utronix.de>,
Vlastimil Babka <vbabka@...e.cz>, linux-kernel@...r.kernel.org,
x86@...nel.org
Subject: Re: [RFC PATCH 1/5] mm: intorduce __GFP_UNMAPPED and unmapped_alloc()
On Thu, May 18, 2023 at 12:03:03PM -0700, Song Liu wrote:
> > Second, bitmap based solution cannot handle > 2MiB allocation cleanly,
> > while tree based solution can. For example, if a big driver uses 3MiB, the
> > tree based allocator can allocate 4MiB for it, and use the rest 1MiB for
> > smaller allocations.
2 MB is just a good default size for the initial unmapped allocations if
the main consideration is avoiding mapping fragmentation. The approach
can work with larger allocations if required.
Powered by blists - more mailing lists