From mboxrd@z Thu Jan 1 00:00:00 1970 From: Adolf Belka To: development@lists.ipfire.org Subject: Re: Problems still with xfs Date: Fri, 09 Feb 2024 14:11:58 +0100 Message-ID: <1167237b-01fc-427a-868e-ae06951622f2@ipfire.org> In-Reply-To: <1cd61387-6d99-4428-8f84-a08d7b970363@ipfire.org> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============6546560040643138295==" List-Id: --===============6546560040643138295== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Hi Michael, On 09/02/2024 13:20, Adolf Belka wrote: > Hi Michael, > > On 09/02/2024 13:00, Michael Tremer wrote: >> Hello Adolf, >> >> Thanks for raising this. I had a look at this, but I believe it is a false= alarm :) > :+1: >> >>> On 8 Feb 2024, at 12:23, Adolf Belka wrote: >>> >>> Hi Michael and Arne, >>> >>> I decided to build a vm testbed clone that would have an xfs filesystem. >>> >>> I did the installation from Core Update 182 downloaded from the website. = After the reboot during the installation the screen went blank and nothing mo= re happened. >> >> This is correct. Looking at the installation log, GRUB fails to install an= d shows a message. Sadly we don=E2=80=99t properly catch that on the installe= r. >> >>> I then checked the sha256 hash for the website version with the one in th= e nightly build for Core Update 182 with the grub reversion and they are diff= erent. >> >> Yes, this is also correct. We rebuilt the update with GRUB 2.06 after seei= ng some problems with GRUB 2.12-rc1. But we actually only changed the updater= and did not change the installation images. >> >> This has two reasons: Mainly changing the images is a pain. We have ~30 mi= rrors and getting them all up to date is too difficult. People might download= parts of the file from different servers which results in an utterly broken = image. > > I now remember you mentioning this for an earlier change but I had forgotte= n that. > Can understand that it would be a right pita. > >> >> The other one is that the problem has an easy workaround: Don=E2=80=99t us= e XFS. So installation of IPFire is possible. If you really want XFS, install= the previous version and update. >> >>> I then built a system with xfs filesystem using the CU182 from the nightl= y build with the grub reversion but it still failed at the reboot stage of th= e installation. >> >> That should actually work. I just tested the current master image (https:/= /nightly.ipfire.org/master/2024-02-02%2007:52:09%20+0000-8c43d148/x86_64/) an= d this boots fine on XFS. > Ah. That might be where my problem was. I used the CU182 nightly build. > > https://nightly.ipfire.org/core182/2024-01-10%2021:49:02%20+0100-c903d67c/x= 86_64/ > > The changelog for that mentioned the grub revert commit so I thought that w= as the one I should use. > > Is master now the 182 version? I thought it was still the 183 testing one? > > Anyway, I will test out the one from your link to confirm. > So I tested out the link for the master version you provided but that gave th= e CU183 iso. xfs file system built without any problems with CU183. I then tried the CU182 version again from the nightly and it failed again. Th= e log for that build shows it has grub-2.06 so the reversion was definitely a= pplied. As it is a vm build that I am doing I can't access the logs from the install = to see what is going wrong. However it is probably best to leave everything as it is. Most people will us= e the download site for the iso's and can download CU181 and then update to C= U182, if they want to use xfs. CU183 will work fine with xfs so grub-2.12 has resolved whatever the problem = was. Regards, Adolf. >> >> >>> >>> I then did an xfs build using the CU181 website download version and that= built without any problems. >>> >>> I then did a pakfire update from CU181 to CU182 and that went without any= problems, including the reboot. >>> >>> Not sure what is going on but there seems to still be a problem with doin= g a fresh installation with xfs using either the website or nightly build ver= sions of CU182. >> >> That might be, but it should be fine from Core Update 183. > I hadn't actually tested out the Core Update version, so I will also try an= d confirm that. > > Sorry for any noise. I just thought it was worth having an xfs based system= on my testbed so at least I can check the update and booting when a Testing = release is done. > > Regards, > Adolf. >> >> -Michael >> >>> >>> Regards, >>> >>> Adolf. >>> >> --===============6546560040643138295==--