From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: MIME-Version: 1.0 References: <20191030124431.11242-1-cgxu519@mykernel.net> <16e204de70e.cefd69461771.2205150443916624303@mykernel.net> <16e314ad3bc.f4c363d96385.3761437052169638038@mykernel.net> In-Reply-To: <16e314ad3bc.f4c363d96385.3761437052169638038@mykernel.net> From: Amir Goldstein Date: Sun, 3 Nov 2019 16:46:05 +0200 Message-ID: Subject: Re: [PATCH v2] ovl: improving copy-up efficiency for big sparse file Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable To: Chengguang Xu Cc: Miklos Szeredi , overlayfs , Vivek Goyal List-ID: On Sun, Nov 3, 2019 at 2:43 PM Chengguang Xu wrote: > > ---- =E5=9C=A8 =E6=98=9F=E6=9C=9F=E5=9B=9B, 2019-10-31 14:53:15 Amir Gol= dstein =E6=92=B0=E5=86=99 ---- > > > > Yes, overlayfs does not comply with this "posix"' test. > > > > This is why it was removed from the auto and quick groups. > > > > > > So I'm curious what is the purpose for the test? > > > > > > > This is a POSIX compliance test. > > It is meant to "remind" us that this behavior is not POSIX compliant > > and that we should fix it one day... > > A bit controversial to have a test like this without a roadmap > > when it is going to be fixed in xfstests, but it's there. > > I haven't checked carefully for the detail but It seems feasible if we c= opy-up lower file during mmap regardless of ro/rw mode. > Is it acceptable by slightly changing copy-up assumption to fulfill POSI= X compliance? Or we just wait for a better solution? > That was attempted in the past. It's complicated ;-) Cheers, Amir.