On Fri, Jan 24, 2020 at 01:07:27PM +1100, Dave Chinner wrote: > On Wed, Jan 22, 2020 at 11:43:07PM -0800, Darrick J. Wong wrote: > > From: Darrick J. Wong <darrick.wong@xxxxxxxxxx> > > > > Instead of passing __func__ to the error reporting function, let's use > > the return address builtins so that the messages actually tell you which > > higher level function called the buffer functions. This was previously > > true for the xfs_buf_read callers, but not for the xfs_trans_read_buf > > callers. > > > > Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx> > > --- > > fs/xfs/xfs_buf.c | 12 +++++++----- > > fs/xfs/xfs_buf.h | 7 ++++--- > > fs/xfs/xfs_buf_item.c | 2 +- > > fs/xfs/xfs_log_recover.c | 4 ++-- > > fs/xfs/xfs_trans_buf.c | 5 +++-- > > 5 files changed, 17 insertions(+), 13 deletions(-) > > Makes sense, but I have a concern that this series now has added two > new parameters to the buffer read functions. Perhaps we should consider > wrapping this all up in an args structure? Yes. Next cycle I'll try to cough up a patchset to turn this into an args structure so that callers can pass additional things like health reporting breadcrumbs to the verifiers so that we can update that status without having to explode a ton of error handling boilerplate all over the place. --D > That's a separate piece > of work, not for this patchset. > > So far this patch, > > Reviewed-by: Dave Chinner <dchinner@xxxxxxxxxx> > -- > Dave Chinner > david@xxxxxxxxxxxxx
