On Tue, Dec 10, 2024 at 07:29:39AM -0500, Brian Foster wrote: > This patch appears to cause a regression with CONFIG_XFS_RT disabled. On > generic/034 or generic/039, I see a mount failure and hang with the > following in dmesg: > > [ 522.627881] XFS (dm-2): Mounting V5 Filesystem c984a654-8653-430a-9042-b59f26132eb6 > [ 522.676496] XFS (dm-2): Starting recovery (logdev: internal) > [ 522.693551] XFS (dm-2): log mount/recovery failed: error -95 > [ 522.701508] XFS (dm-2): log mount failed > > It seems to be related to the xfs_update_last_rtgroup_size() definition > above in that the issue goes away if I replace the error code with zero, > but I've not looked further into it than that. I'm also not clear on if > the hang behavior is related or just a separate, latent issue triggered > by the injected error. So without CONFIG_XFS_RT we obviously should not update rgcounts, but should also fail the mount earlier if there are RGs. Turns out that non-rtg file systems have a fake RTG if they have a rt subvolume, and the count is always set to 1. So yes, this should just return 0 and your fix is correct.
