svn commit: r220412 - in head: share/man/man4 sys/cam/ata

Garrett Cooper yanegomi at gmail.com
Thu Apr 7 22:02:24 UTC 2011


On Thu, Apr 7, 2011 at 2:23 PM, Warner Losh <imp at bsdimp.com> wrote:
>
> On Apr 7, 2011, at 3:00 PM, Garrett Cooper wrote:
>
>> On Thu, Apr 7, 2011 at 12:03 PM, Alexander Best <arundel at freebsd.org> wrote:
>>> On Thu Apr  7 11, Alexander Motin wrote:
>>>> Author: mav
>>>> Date: Thu Apr  7 08:17:53 2011
>>>> New Revision: 220412
>>>> URL: http://svn.freebsd.org/changeset/base/220412
>>>>
>>>> Log:
>>>>   Make ada(4) driver to control device write cache, same as ata(4) does.
>>>>   Add kern.cam.ada.write_cache sysctl/tunable to control it alike hw.ata.wc.
>>>
>>> how hard would it be to support per device sysctls/tunables? i'd really like to
>>> do:
>>>
>>> kern.cam.ada.0.write_cache=0 (root fs)
>>> kern.cam.ada.1.write_cache=1 (/usr, /var, etc.)
>>
>> Does it really make sense to turn on write caching for one drive and
>> not the other(s)?
>
> Think about /usr/obj or /tmp and ask that question again.  Or any filesystem that's mounted that you don't really care about the contents of across a power cycle.  If you have to recreate it, that's OK.   In those cases, you may want the speed increase over safety that write_cache gives you.  Or maybe you have a drive that's doing write caching correctly and one that doesn't.

    Alex, Warner: thanks for the practical example; I hadn't really
considered that :).
Cheers!
-Garrett


More information about the svn-src-head mailing list