RfC: fueword(9) and casueword(9)

[Konstantin Belousov]

On Sat, Oct 25, 2014 at 09:43:44PM +0200, Jilles Tjoelker wrote:
> On Tue, Oct 21, 2014 at 07:23:06PM +0300, Konstantin Belousov wrote:
> > I prefer to not complicate the fetch(9) KPI due to the mistakes in the
> > umtx structures definitions.  I think that it is bug to mark the lock
> > words with volatile.  I want the fueword(9) interface to be as much
> > similar to fuword(9), in particular, volatile seems to be not needed.
> 
> > Below is the updated patch, together with the bug fix noted by Eric.
> 
> Hmm, consider returning an error number (that is, EFAULT) instead of -1
> on failure. This is somewhat like priv_check() which returns EPERM on
> failure, and probably reduces confusion if the return value is assigned
> to a variable named "error".
I did considered EFAULT.  The KPI becomes too inconsistent if comparing
fueword with fuword or suword.

If you reference rv in kern_umtx.c, returning EFAULT for fueword or
casueword would not allow to use error directly.  The reason is the
complicated logic of handling the suspend requests.  Basically, non-errors
from userspace access must not obliterate the errors from previous
calls to check_suspend().  See r270345 for example of the bug from
earlier pass over userspace access in kern_umtx.c.  There are still
more such bugs left in the patch, Peter and I chasing them.

> 
> In share/man/man9/Makefile, MLINKS are still created for fuebyte.9 and
> fueword16.9.
Fixed, thank you.

> 
> "successful" is consistently misspelled "successfull".
I corrected all occurences of successfull in the patch, but it is systematic
in the tree.

pooma% git grep -e successfull -- | grep -v successfully | wc -l
      30