Update to the ZFS detection logic to be stable. If any
component of the I/O is ZFS treat the entire I/O as ZFS
HPE-bug-id: LUS-12175
Fixes:
07a1272233 ("LU-17525 llite: unaligned DIO interop page alignment")
Signed-off-by: Shaun Tancheff <shaun.tancheff@hpe.com>
Change-Id: Ie9d5c8af373cec96855f81f5330e174ed71f370a
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/55895
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
* avaliable mirror for write.
*/
io->ci_ndelay = !(iot == CIT_WRITE);
+ io->ci_target_is_zfs = 0;
/* unaligned DIO has compat issues with some older servers, but we find
* out if there are such servers while setting up the IO, so it starts
* out allowed
CL_IO_SLICE_CLEAN(oio, oi_cl);
cl_io_slice_add(io, &oio->oi_cl, obj, &osc_io_ops);
- cl_io_top(io)->ci_target_is_zfs = exp_connect_target_is_zfs(exp);
+ if (exp_connect_target_is_zfs(exp))
+ cl_io_top(io)->ci_target_is_zfs = 1;
if (!exp_connect_unaligned_dio(exp))
cl_io_top(io)->ci_allow_unaligned_dio = false;