> On Oct. 14, 2015, 5:36 a.m., Matthew Ahrens wrote: > > usr/src/uts/common/fs/zfs/dmu.c, lines 444-447 > > <https://reviews.csiden.org/r/259/diff/1/?file=17810#file17810line444> > > > > I don't see that DNODE_IS_CACHEABLE has ever existed in illumos. > > Alexander Motin wrote: > I've added it to sys/dnode.h. If you have better idea -- I am open for > propositions. So far I've found only ones for dbuf and objset, but none of > them is really applicable here.
Ah, sorry, for some reason I didn't see the definition in dnode.h. This macro name is fine with me. > On Oct. 14, 2015, 5:36 a.m., Matthew Ahrens wrote: > > usr/src/uts/common/fs/zfs/dmu_zfetch.c, lines 289-290 > > <https://reviews.csiden.org/r/259/diff/1/?file=17811#file17811line289> > > > > I think I understand why you are rounding up the start -- if the start > > is misaligned (not at the beginning of the indirect block) we would have > > alread prefetched this indirect last time. > > > > Why do you round up the end? It seems like you would want to end on > > the indirect that contains the ending level-0 blkid (zs_ibf_blkid). > > Alexander Motin wrote: > You may see that loop below does not actually fetch ipf_iend block, > stopping just before it. So to me it looks consistent. ah, I see. > On Oct. 14, 2015, 5:36 a.m., Matthew Ahrens wrote: > > usr/src/uts/common/fs/zfs/sys/dmu_zfetch.h, line 46 > > <https://reviews.csiden.org/r/259/diff/1/?file=17812#file17812line46> > > > > it would be helpful to specify that this is the blkid of the level-0 > > block whose level-1 indirect we will prefetch (as opposed to the blkid of > > the next level-1 indirect to prefetch) > > Alexander Motin wrote: > Please propose wording that you, as native speaker, would prefer. And > that would better not tool whole paragraph. :) How about: /* next level-0 block whose indirect to prefetch */ - Matthew ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.csiden.org/r/259/#review833 ----------------------------------------------------------- On Oct. 14, 2015, 6:37 a.m., Alexander Motin wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.csiden.org/r/259/ > ----------------------------------------------------------- > > (Updated Oct. 14, 2015, 6:37 a.m.) > > > Review request for OpenZFS Developer Mailing List and Christopher Siden. > > > Bugs: 6322 > https://www.illumos.org/issues/6322 > > > Repository: illumos-gate > > > Description > ------- > > 6322 ZFS indirection tables prefetch > > > Diffs > ----- > > usr/src/uts/common/fs/zfs/sys/dmu_zfetch.h > 6f61198ebc4528f58761f56a11610cb2fbefadd3 > usr/src/uts/common/fs/zfs/dmu.c 5b24268af1a813c153c75e8531cb3a6e23baa406 > usr/src/uts/common/fs/zfs/dbuf.c 8b0c71db5c86fab2219699f4e3b2eccf00de545c > usr/src/uts/common/fs/zfs/sys/dnode.h > 69cc54dc272b143b085f897b956bb9cdd8c09bf3 > usr/src/uts/common/fs/zfs/dmu_zfetch.c > f2cdf863d83c9ecb627a3b566c4c09553e4b9666 > > Diff: https://reviews.csiden.org/r/259/diff/ > > > Testing > ------- > > Tested on FreeBSD head system. > > > Thanks, > > Alexander Motin > >
_______________________________________________ developer mailing list [email protected] http://lists.open-zfs.org/mailman/listinfo/developer
