Steven Hartland
2013-08-19 00:23:55 UTC
Few things I've noticed:-
1. in vdev_geom.c is there a reason you use:
+ *physical_ashift = 0;
+ if (pp->stripesize)
+ *physical_ashift = highbit(pp->stripesize) - 1;
Instead of:
+ *physical_ashift = highbit(MAX(pp->stripesize, SPA_MINBLOCKSIZE)) - 1;
2. sysctl_vfs_zfs_max_auto_ashift should also limit the min to >
SPA_MINBLOCKSHIFT
3. sysctl_vfs_zfs_max_auto_ashift should error if the value is out
of range instead of clamping it.
One thing we did here, which would be good to see in this patch, is to add
an overall min system ashift as thie enables admins to configure pools
to be compatible with future disks they are likely to use e.g. min ashift
12 (4k compatible). This could be left at 9 by default for max compatibility
but personally I'd suggest 12.
Regard
Steve
----- Original Message -----
From: "Justin T. Gibbs" <***@FreeBSD.org>
To: <zfs-***@freebsd.org>; <***@lists.illumos.org>
Cc: "Richard Yao" <***@gentoo.org>
Sent: Monday, August 19, 2013 12:42 AM
Subject: Updated ashift optimization patch
Today I re-merged all of Spectra's ashift related changes into FreeBSD/head. Here's the current state of the patch. See the top
of the patch for proposed commit message/justification. The majority of it is ZFS platform agnostic.
Later this week I'll get a stock FreeBSD/head system running and run the test suites in order to verify that I haven't botched
anything in the merge. I'll also see about merging it to one of Will's Illumos VMs so I can generate a webrev. In the mean time,
comments, concerns, improvements welcome.
--
Justin
--------------------------------------------------------------------------------
This e.mail is private and confidential between Multiplay (UK) Ltd. and the person or entity to whom it is addressed. In the event of misdirection, the recipient is prohibited from using, copying, printing or otherwise disseminating it or any information contained in it.
In the event of misdirection, illegible or incomplete transmission please telephone +44 845 868 1337
or return the E.mail to ***@multiplay.co.uk.
1. in vdev_geom.c is there a reason you use:
+ *physical_ashift = 0;
+ if (pp->stripesize)
+ *physical_ashift = highbit(pp->stripesize) - 1;
Instead of:
+ *physical_ashift = highbit(MAX(pp->stripesize, SPA_MINBLOCKSIZE)) - 1;
2. sysctl_vfs_zfs_max_auto_ashift should also limit the min to >
SPA_MINBLOCKSHIFT
3. sysctl_vfs_zfs_max_auto_ashift should error if the value is out
of range instead of clamping it.
One thing we did here, which would be good to see in this patch, is to add
an overall min system ashift as thie enables admins to configure pools
to be compatible with future disks they are likely to use e.g. min ashift
12 (4k compatible). This could be left at 9 by default for max compatibility
but personally I'd suggest 12.
Regard
Steve
----- Original Message -----
From: "Justin T. Gibbs" <***@FreeBSD.org>
To: <zfs-***@freebsd.org>; <***@lists.illumos.org>
Cc: "Richard Yao" <***@gentoo.org>
Sent: Monday, August 19, 2013 12:42 AM
Subject: Updated ashift optimization patch
Today I re-merged all of Spectra's ashift related changes into FreeBSD/head. Here's the current state of the patch. See the top
of the patch for proposed commit message/justification. The majority of it is ZFS platform agnostic.
Later this week I'll get a stock FreeBSD/head system running and run the test suites in order to verify that I haven't botched
anything in the merge. I'll also see about merging it to one of Will's Illumos VMs so I can generate a webrev. In the mean time,
comments, concerns, improvements welcome.
--
Justin
--------------------------------------------------------------------------------
_______________________________________________
http://lists.freebsd.org/mailman/listinfo/zfs-devel
================================================http://lists.freebsd.org/mailman/listinfo/zfs-devel
This e.mail is private and confidential between Multiplay (UK) Ltd. and the person or entity to whom it is addressed. In the event of misdirection, the recipient is prohibited from using, copying, printing or otherwise disseminating it or any information contained in it.
In the event of misdirection, illegible or incomplete transmission please telephone +44 845 868 1337
or return the E.mail to ***@multiplay.co.uk.