Hi, This patchset contains changes to linux and to xfstests to address a reoccuring panic in xfs_bmap_extents_to_btree. The RFC was discussed here: https://lore.kernel.org/linux-xfs/cover.1718232004.git.kjlx@xxxxxxxxxxxxxxxxxx/T/#t The kernel changes modify how the AGFL reservation is calculated when a filesystem is mounted. This is also pushed into the in-core per-AG structures to ensure that they do not consume the additional space reserved by this change. Additionally, this includes a pair of xfstest patches. The first introduces a test that triggers the problem we're trying to fix, as xfs/608. The second is a modification to xfs/306, which started failing because the increased space that is reserved by these changes is above the global reserve limit that this test intentionally lowers. The second patch increases this limit by one block. If this seems wrong, I'm happy to debug further. The change was based upon the assumption that artifically lowered global reservation limits had to be cognizant of the per-AG limits. -K
