A typical bug around VOP_GETPAGES?

Yar Tikhiy yar at comp.chem.msu.su
Mon Aug 30 02:20:10 PDT 2004

Hi folks,

Users' complaints about HFS for FreeBSD not compiling again encouraged
me to hack up a yet another patchlevel allowing for recent changes
in our beloved OS :-)

This time fsx started to fail upon a mmapped access near the end
of an HFS file with diagnostics that non-zero data were found beyond
EOF (which is against POSIX.)  I'm confused by this.  HFS relies
on the default VOP_GETPAGES handler and hence I used to believe
that the kernel should just bread() the last block of the file,
copy the appropriate part of it to the process' page, and zero out
the rest of the page.   Fsx found no other problems in a reasonable
time, which meant plain vanilla reading from (and writing to) the
FS worked OK.

Perhaps anyone has a quick idea where to look for problematic code?


More information about the freebsd-fs mailing list