Dan Aloni reported a problem with the way NFSD's READ implementation
deals with the very upper end of file sizes, and I got interested in
how some of the other operations handled it. I found some issues,
and have started a (growing) pile of patches to deal with them.
Since at least the SETATTR case appears to cause a crash on some
filesystems, I think several of these are 5.17-rc fodder (i.e.,
priority bug fixes).
I've already posted an update to the pynfs CMT4 test. A similar
pynfs test for out-of-range READ is forthcoming.
Changes since v2:
- Addressed concerns with "Fix NFSv3 SETATTR/CREATE's handling of
large file sizes"
- Added fix for READ underflow, as initially reported; series is now
complete
- Trace points now report wire input values before type cast
Changes since RFC:
- Series reordered so priority fixes come first
- Setattr size check is now in a common function
- Patch descriptions clarified
---
Chuck Lever (7):
NFSD: Fix the behavior of READ near OFFSET_MAX
NFSD: Fix ia_size underflow
NFSD: Fix NFSv3 SETATTR/CREATE's handling of large file sizes
NFSD: Clamp WRITE offsets
NFSD: COMMIT operations must not return NFS?ERR_INVAL
NFSD: Fix offset type in I/O trace points
NFSD: Deprecate NFS_OFFSET_MAX
fs/nfsd/nfs3proc.c | 19 ++++++++-------
fs/nfsd/nfs3xdr.c | 4 ++--
fs/nfsd/nfs4proc.c | 13 +++++++----
fs/nfsd/nfs4xdr.c | 10 +++-----
fs/nfsd/trace.h | 14 +++++------
fs/nfsd/vfs.c | 57 +++++++++++++++++++++++++++++++--------------
fs/nfsd/vfs.h | 4 ++--
include/linux/nfs.h | 8 -------
8 files changed, 74 insertions(+), 55 deletions(-)
--
Chuck Lever
