Date: Tue, 20 Aug 2002 12:01:29 +0200 From: Matthias Andree <matthias.andree@gmx.de> To: freebsd-stable@freebsd.org Subject: Re: root (/) not soft-updates by default ? Message-ID: <m3it25x30m.fsf@merlin.emma.line.org> In-Reply-To: <200208191604.g7JG4DOU077232@apollo.backplane.com> (Matthew Dillon's message of "Mon, 19 Aug 2002 09:04:13 -0700 (PDT)") References: <20020819144928.GA6628@nebula.wanadoo.fr> <20020819105452.A14530@blackhelicopters.org> <m3ofbylv71.fsf@merlin.emma.line.org> <200208191604.g7JG4DOU077232@apollo.backplane.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Matthew Dillon <dillon@apollo.backplane.com> writes: > This conversation is moot, since modern IDE drives (and apparently > seeping into SCSI drives as well) do track-at-once writes to the > physical media, there is no ordering guarentee no matter what you > do, and IDE performance with write caching turned off sucks so > badly you pretty much *have* to turn it on if tags are not > supported or suffer penalties so severe that the last time we > tried it our user community was up in arms over the result. Could you reveal the source of this information? URL, product manual, something? It contradicts what I have read in several hard drive manuals (that claim that at most one block can be lost as the power fails with the write cache off), and in fact all my IDE drives suffer severe performance drops (50 to 70%) on linear writes already as I turn write caches off -- originating from the lock-step behaviour of "perform a write operation" - "wait for data to hit permanent storage" - "acknowledge write operation". Keep in mind that tags won't to for PIO, and if there was a drive that reordered writes, but did not support dma queued and flush cache command, all bets to consistency would be off. I checked again for Maxtor D740X-6L and IBM Deskstar 120GXP, and either this "track write" is hidden deeply or the drives do the right thing with the write cache off. > Since very little on the root partition is modified once a system has > come up, not enabling softupdates on root is not a big deal. Indeed. -- Matthias Andree To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?m3it25x30m.fsf>