Messages in this thread | | | From | Bartlomiej Zolnierkiewicz <> | Subject | Re: [PATCH 0/32] ide-tape redux v1 | Date | Wed, 30 Jan 2008 01:29:55 +0100 |
| |
On Monday 28 January 2008, Borislav Petkov wrote: > Hi Bart, > > [...] > > > > the BKL in idetape_write_release() with finer-grained locking etc, probably also > > > some pipeline improvements, removal of OnStream support, etc. but that'll come > > > later. > > > > On-Stream support has been long gone but it seems that deprecation > > warning etc. managed to survive. > > > > w.r.t. to the pipeline-mode: it should be pipelined into /dev/null > > > > rationale: > > - it is _very_ complex > > - causes errors to be deferred till the next user-space access > > - direct I/O using blk_rq_map_user() will offer superior performance > > > > the only question is whether to remove it... > > Well, on the one hand, since the driver is only being maintained we should not > remove code that works. Also, i don't know how many users ide-tape really has > but, would it be worth the trouble at all? Because if nobody's using it, we > could just as well pipe the whole thing into /dev/null.. On the other hand, the
This may be the other alternative... [ there is always libata PATA... ]
If you want to give ide-tape removal a try, go ahead (I suggest starting with adding warning printk() and keeping patch in -mm for some time)...
> pipelining part _is_ kinda big and, right, it is not that straightfoward to > look at it and know what it actually does - it truly is a student project :)
I have pipelining code figured out to some degree but reworking it is a rather low-prio on my TODO list...
[...]
> Anyway, resending #23 to you in a private mail.
Thanks.
Bart
| |