[Bug 264196] sysutils/fusefs-ntfs: read(2)ing big files in blocks of 32768 bytes and smaller fails with EINVAL

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=264196

            Bug ID: 264196
           Summary: sysutils/fusefs-ntfs: read(2)ing big files in blocks
                    of 32768 bytes and smaller fails with EINVAL
           Product: Ports & Packages
           Version: Latest
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Only Me
          Priority: ---
         Component: Individual Port(s)
          Assignee: ports-bugs@FreeBSD.org
          Reporter: danfe@FreeBSD.org
                CC: asomers@FreeBSD.org, freebsd@dussan.org
             Flags: maintainer-feedback?(freebsd@dussan.org)
                CC: freebsd@dussan.org
 Attachment #234168 text/plain
         mime type:

Created attachment 234168
  --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=234168&action=edit
Demo program that does the read(2)

With the recent 14.0-CURRENT (n255230) and fusefs-ntfs-2017.3.23 (UBLIO=off),
I'm frequently getting errors when copying not-very-small files with scp(1) to
remote machines, but not with rsync(1); cat(1)ing file e.g. to /dev/null also
fails.  Running those programs under truss(1) reveals interesting bug: reading
the files on Windows 10 formatted NTFS with block of size 32768 and smaller
(stat(2) returns optimal size of 4096 bytes, which both scp(1) and cat(1) use)
fails after several blocks; rsync(1) uses 262144 and does not fail.  Even more
interestingly, it does not fail on first couple of blocks and on previously
read ones, i.e. once the file had been read, it can be read again with smaller
block size.  I've written a program to demonstrate this (attached):

$ ls -l xorgproto-2021.5.tbz
-rw-r--r--  1 danfe  wheel  245998 21 янв.  14:11 xorgproto-2021.5.tbz
$ /tmp/a.out xorgproto-2021.5.tbz 32768
using i/o size 32768 (optimal 4096)
read: Invalid argument
1 reads were successful

$ /tmp/a.out xorgproto-2021.5.tbz 32768
using i/o size 32768 (optimal 4096)
read: Invalid argument
3 reads were successful

$ /tmp/a.out xorgproto-2021.5.tbz 32768
using i/o size 32768 (optimal 4096)

$ /tmp/a.out xorgproto-2021.5.tbz -1
using i/o size 4096 (optimal 4096)

I'm not sure if the bug is in the port or fusefs(5) implementation, so I'm
CC'ing Alan in case he might have a clue.

-- 
You are receiving this mail because:
You are the assignee for the bug.