On Mon, Feb 24, 2020 at 02:27:37PM -0800, Darrick J. Wong wrote: > On Mon, Feb 24, 2020 at 02:21:18PM -0800, Christoph Hellwig wrote: > > On Mon, Feb 24, 2020 at 02:19:31PM -0800, Darrick J. Wong wrote: > > > > > Signed-off-by: Christoph Hellwig <hch@xxxxxx> > > > > > > > > Any chance we can pick this up for 5.6 to unbreak arm OABI? > > > > > > Yeah, I can do that. Is there a Fixes: tag that goes with this? > > > > I'm not sure what to add. I think the problem itself has actually > > always been around since adding v5 fs support. But the build break > > was only caused by the addition of the BUILD_BUG_ON. > > Hmm. That's tricky, since in theory this should go all the way back to > the introduction of the v5 format in 3.x, but that's going to require > explicit backporting to get past all the reorganization and other things > that have happened. We might just have to hand-backport it to the > stable kernels seeing how the macro name change will probably cause all > sorts of problems with AI backports. :/ So which fixes tag do you want? Or feel free to just add the one you feel fits best. > > > Also, will you have a chance to respin the last patch for 5.7? > > > > Last patch in this series? > > Yes. From the discussion of patch 6/6, > > "+ __xfs_sb_from_disk(&sb, bp->b_addr, false); > > "why not dsb here > > "Yes, this should just pass dsb." Oh. I've actually had the respun branch on my box since a day after that comment. But I think it doesn't make sense until the fix in patch one is in the baseline tree, given how many outstanding patch series we have.
