I hit two bugs in ext4 ea_inode feature when testing Darrick's new test case[1] (which is still in review & revision process). The first bug is premature ENOSPC, and the second one is a refcount corruption. I've tested the fixes with fstests and xattr related syscall tests in LTP, and tests all passed. Thanks, Eryu [1] [PATCH] generic: test XATTR_REPLACE doesn't take the fs down https://www.spinics.net/lists/linux-xfs/msg17394.html Eryu Guan (2): ext4: use XATTR_CREATE in ext4_initxattrs() ext4: use raw i_version value for ea_inode fs/ext4/inode.c | 24 ++++++++++++++++++++++-- fs/ext4/xattr_security.c | 2 +- 2 files changed, 23 insertions(+), 3 deletions(-) -- 2.14.3
