Finally done with one big functionality I was trying to implement this weekend, an In-Core extent tree for Inodes in XFS.
We didn't had any centralized functionality where we can access and query extents
we read large files to check their MD5 hash to see if we traverse correctly
2
43
I don’t wanna be that person but oh well happy poison air Diwali.
I just we could celebrate this amazing festival pollution free :’)
Some more bug fixes on extended attributes, I basically found that lookup for attribute name returned entry when name of one is prefix of another (eg: "ATTR1" and "ATTR123") due to incorrect name comparison.
Well that is fixed now :)
Well Leaf and Node xattrs format works again. I found that there was regression introduced with re-ordering of on-disk leaf and node attributes metadata struct, was a quick fix
In the example below we read node attributes having 820 entries each with 8 bytes of data.
Haiku xfs driver just got faster!
Implemented entry cache for the xfs driver so that same Inode can have very fast lookup (directly from cache) rather than from disk on every inode read.
Fixed SMAP violation for ext2 and xfs file_system driver in haiku. Looks like there was some change in kernel and only bfs driver was properly upgraded to use user_memcpy.
I previously fixed that for xfs while reading file contents but this one is for reading extended attributes.
Haiku XFS driver now supports large extents counter, this was one of the last major feature that read only driver should support to be able to read latest xfs formatted disks.
I should now begin to make driver read support as performant and as bug free as possible, lets see :)
Holy shit dude….
Apparently this guy wants AI to make love to his wife since he’s so busy working with AI.
1
52











