From mboxrd@z Thu Jan 1 00:00:00 1970 From: Neil Brown Subject: RE: [PATCH 3/6] Enable create array with write journal (--write-journal DEVICE). Date: Mon, 19 Oct 2015 13:42:08 +1100 Message-ID: <87fv17lgu7.fsf@notabene.neil.brown.name> References: <1440804426-1461372-1-git-send-email-songliubraving@fb.com> <1440804426-1461372-4-git-send-email-songliubraving@fb.com> Mime-Version: 1.0 Content-Type: multipart/signed; boundary="=-=-="; micalg=pgp-sha256; protocol="application/pgp-signature" Return-path: In-Reply-To: Sender: linux-raid-owner@vger.kernel.org To: Song Liu , Dan Williams Cc: linux-raid , Shaohua Li , Christoph Hellwig List-Id: linux-raid.ids --=-=-= Content-Type: text/plain Content-Transfer-Encoding: quoted-printable Song Liu writes: >>=20 >> Hmm, perhaps this should instead follow the same command line mechanics >> as "--write-mostly" whereby all the devices listed after that flag take = on that >> role. Also, if we anticipate more roles I wonder if the option should b= e "-- >> role=3D" that assigns the given role to the device(s) that follow. >>=20 >> Thoughts? >>=20 >> I'm looking at this from the perspective of reviving the SRT patches [1]= where >> a role for a read/write cache device is needed. Where a read/write-cach= e is >> different than a journal. >>=20 >> [1]: >> https://urldefense.proofpoint.com/v1/url?u=3Dhttps://lwn.net/Articles/59= 661 >> 4/&k=3DZVNjlDMF0FElm4dQtryO4A%3D%3D%0A&r=3DJ3jGe56dIPfS5TN6DM82U >> YbbeR1j2viaiSJI40tv6lE%3D%0A&m=3Ddv5znFd%2BTZdHZaSKjzBVqJbdtNBALR0 >> RgzhZsOmsWAU%3D%0A&s=3D80c812981ff98264cf1bb2b79032863d4e33ba100 >> 093d41956eab3fbd33b5635 > > I think --write-mostly is used as _hint_ to RAID 1 that differentiates SS= D from=20 > HDD. While --write-journal turns on "MD_FEATURE_JOURNAL" and specifies=20 > journal device. We should not need more than 1 journal device. In this se= nse,=20 > I think current --write-journal works well... =2D-write-mostly isn't a hint. It is a clear statement that these devices should only be read from if there is no other option (the original justification had nothing to do with SSDs). However a write journal is a very different thing to the other devices. For example it is not counted in the --raid-disks and --spare-disks where as all other devices (whether --write-mostly or not) are. If anything --write-journal is more like a --bitmap=3D/some/file than =2D-write-mostly. So, I'm happy for --write-journal to stay as it is. Thanks, NeilBrown --=-=-= Content-Type: application/pgp-signature; name="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBCAAGBQJWJFiAAAoJEDnsnt1WYoG59HcQAKdNs6dpvgpjY7ZLrAXYprjq VoEMzzpOUL/eD9q4qZbD5QGpHPXEBZsSKLBv9CdL5zsJ3RYJwhhRo5TMLxsTk9Td 09oVIaV+x0ST4CTq8xPHjpbf18WlgwMRDi+IYyneKjBGQBbpALvrUlRxtn1g1KbN exvQJQhmnBjjEva1AQoCZ7aGharO1nOlEI7DMTqPFVxd59cPc7J3xCuzFGM0NBBW qXxYdECLhW9sUjBzkSG7SMkQJoQIrjNljQ0rUfHRktDuC9mfaOz3EPRZJFXT6ULE ZC2ZuG9Lkw4wXxTYrNLRallfTLVsmk9gIwpUqxyIYfPcwkWVkp3SROZLr8oCYi09 ONYSkVfXTd8yG/q1zQprc5VteMpdA3jxtBZcCBVxOfdMpxeQn98FPgqAgGqXn36Y aVIaBt0Ymr96++MYOSi3MRehsSqA3ws9fwpVDoXRm5bajFHZtaWIAC5XUsRACufh I+618fsqw2pW5OHg/JJKwUaSD2UL2uRceStx6mtmQjfeO4vg4VzhUHZe4+XpZuTZ gGlIpfd26KV8D3I6zXihZIKz0BTbvW8MYRFeMc+1AepQemhgXuVy/S6WbKbABvcB YTEpxI5MvzlkdNpczTaBCdr6NdLLLq70xtpWkTPWLMFpVLk9HJcmO5yoKtYvzwCE JM3QMLvSdJ725ZVma4vL =A3gT -----END PGP SIGNATURE----- --=-=-=--