Date: Fri, 24 Sep 2010 10:05:15 +0300 From: Kostik Belousov <kostikbel@gmail.com> To: Bruce Evans <brde@optusnet.com.au> Cc: freebsd-fs <freebsd-fs@freebsd.org> Subject: Re: Strange behaviour with sappend flag set on ZFS Message-ID: <20100924070515.GD34228@deviant.kiev.zoral.com.ua> In-Reply-To: <20100924083610.B714@delplex.bde.org> References: <201009231938.09548.cal@linu.gs> <20100924083610.B714@delplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--KN5l+BnMqAQyZLvT Content-Type: text/plain; charset=us-ascii Content-Disposition: inline On Fri, Sep 24, 2010 at 10:42:54AM +1000, Bruce Evans wrote: > ffs_write() is also missing the check for read-only mounts. This is subtle. > I think the fs cannot be mounted read-only if ffs_write() is reached -- > even if the file was opened for writing at a time when the mount was > read-write, the file must have been dowgraded to unwriteable of the mount > was successfully downgraded to read-only (takes a forced unmount which > tends to break applications but shouldn't break the file system). The files open for write are effectively closed on rw->ro downgrade. Technically, the vnodes are reclaimed at the time of remount, and still open file descriptors return appropriate deadfs errors on i/o ops, typically ENXIO or EIO. There is a known brokeness with remounts, where the writeable shared mappings of the vnodes are not accounted for the write count of vnode. I have lingering patch for the issue that is not committed for some reasons. --KN5l+BnMqAQyZLvT Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (FreeBSD) iEYEARECAAYFAkycTaoACgkQC3+MBN1Mb4j9rQCbB20GSAE77o7dqqRThCT8B5FA bs4An312pcGy9dGFHTT8iB5CE5s7xAwP =EXZ3 -----END PGP SIGNATURE----- --KN5l+BnMqAQyZLvT--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20100924070515.GD34228>