On Aug 16, 2019, at 5:55 PM, Shehbaz Jaffer <shehbazjaffer007@xxxxxxxxx> wrote: > > Hello, > > I am trying to understand jbd2 checksumming procedure. I reboot ext4 > in the middle of a metadata intensive operation using echo b> > /proc/sysrq-trigger. I see that the journal gets replayed on next > mount using prinks in jbd2/recovery.c: do_one_pass() function. > > I then corrupt intermediate metadata logged on jbd2 and I still see > the journal being replayed without multiple error messages which > should ideally get printed when one of the two following flags - > JBD2_FEATURE_INCOMPAT_CSUM_V2 or JBD2_FEATURE_INCOMPAT_CSUM_V3 are > set. > > I have 2 questions: > > 1. Are the two flags: JBD2_FEATURE_INCOMPAT_CSUM_V2 and > JBD2_FEATURE_INCOMPAT_CSUM_V3 set by default? If not, how do we set > them so that the journal will detect and respond to injected > corruptions? See set_journal_csum_feature_set() for details on how these flags are set. > 2. this is very naive question, but what do compat and incompat > options mean? If flag X in incompat is set, does this mean the feature > does not exist? See description of the compat/incompat/ro_compat fields at: https://ext4.wiki.kernel.org/index.php/Ext4_Disk_Layout#The_Super_Block Cheers, Andreas
Attachment:
signature.asc
Description: Message signed with OpenPGP
