On Fri, Jun 13, 2014 at 10:33:37AM -0400, Douglas Gilbert wrote: >> this looks generally good to me, but I don't think open_cnt and exclude >> have to use atomic_t, as they are only ever modified under open_rel_lock. > > They are read by 'cat /proc/scsi/sg/debug' [in > sg_proc_seq_show_debug()] but transient off-by-1 reports > are not so important. [Prior to locking that block with > read_lock(sfd_lock) that routine sometimes printed wild > results, so some care is required.]
In general 32-bit variables that can be read independly only need the lock for updates. A lock on the read side is important if variables are of a type that can't be atomically read (e.g. 64-bit on 32-bit architectures, or small integers on some old alpha CPUs), or if the values of multiple variables need to be consistent for the reader. >> Can you take a look at the version below? This changes open_cnt to an >> int, exclude to a bool, removes the open_cnt underflow check that >> the VFS takes care for, and streamlines the open path a little bit: > > sg_open() and sg_release() clean-up looks good. The > back-up goto at the end of sg_open() reminds me of > Fortran style :-) More importantly my sg_tst_excl* > tests give this version the thumbs up. > > Acked-by: Douglas Gilbert <[email protected]> Given that this was your patch with minor changes I was planning to put this in under your name if that's fine with you. I'm also looking for another review. -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
