El lun, 7 sept 2026 a las 10:09, x85729201kza (<[email protected]>) escribió: > > > > On 7/09/2026 3:53 pm, Ramiro Aceves wrote: > > Thanks so much for your comments regarding this subject. > > > > I have fsck the ext2 filesystem using Linux and repeated the rsync. > > Now rsync stopped into another file among the Mozilla firefox > > directory that owns trillions of files. > > > > I noticed that when I fixed the filesystem with fsck under Linux it > > output the following "Invalid HTREE directory" message: > > HTREE, according to netbsd, is a hashing in ext2 to speed up searches. > https://github.com/NetBSD/src/blob/55e8a8a44d9e70b9589a9baec5b637d22673df06/sys/ufs/ext2fs/ext2fs_hash.c#L240 > > > > > > sep 06 11:47:53 debian-nuc8i7 systemd-fsck[1293]: /dev/sdb1: Invalid > > HTREE directory inode 38723585 > > (/COPIA_NETBSD_8I7/home/ramiro/.mozilla/firefox/527cpmj0.default-default-2/storage/permanent/chrome/idb/3870112724rsegmnoittet-es.files) > > Jeez, is that systemd now running file I/O as well. Magnificent bloatware > there...
Yes, that is one of the reasons I am a NetBSD fan now! although I cannot remove Debian completely cause I need it for things that do not work in NetBSD ;-) > > > > > > Do you think it is a bug in NetBSD ext2 filesystem implementation? I > > am a total ignorant in this subject, just in case clue persons can > > extract information from this finding. > > > Yes, me too. However, I think your overall challenge is this: ffs is fast has > proper buffer lock handling/optimizations, rsync creates a ton of file I/O > per second (metadata etc), NetBSD's implementation of the VFS called ext2 > cannot handle this. I understand... > > > Anyway, I have excluded from the rsync backup all the offending files > > and it seems to finish just fine. All those files are not important > > for my backup. > > > > Have you perhaps looked at another file system? ext4 can be run using > fuse-ext2. That's far more robust than antique ext2. Geez, maybe even fat32 > could be better? Why not try a small partition or another disk if you have > spare and test them out? I recall in the past having searched about ext4 using fuse and the conclusion I remember is that write support to ext4 filesystems was risky and experimental. The other candidate was exFAT but I discarded cause it has no UNIX permissions, if I recalled correctly. Regards. > > Cheers > > > Not a fix, a workaround but if it works.... > > > > Regards. > >
